Fibertech's visible IPv4 footprint fell from a /22 to a /24; the resilience test is under El Vigía's streets

Summary

• Fibertech's current website advertises unlimited fibre plans of 450 Mbps, 650 Mbps and 1 Gbps in El Vigía, dual-stack IPv4/IPv6, 24-hour monitoring and support, a local address and an external payment portal. These are credible signs of present commercial operation, but not a map of installed fibre, customers, rings, sites or repair resources.
• At 08:00 UTC on 16 July 2026, AS273994 originated one IPv4 /24 and one IPv6 /32. Every collected path for both current prefixes had AS264778, TotalCom Venezuela, immediately before Fibertech. That is one publicly observed logical upstream boundary, not proof of one circuit, one cable or physical single-homing.
• The visible IPv4 estate changed from a historical Cogent-parent /22, last observed in March 2026, to a current provider-aggregatable /24 first observed in June. The intervening absence of an AS273994-originated IPv4 route does not prove a customer IPv4 outage, and the smaller address block does not prove lower bandwidth.
• No reviewed public material establishes Fibertech's fibre ownership, route kilometres, ring closure, separate street ducts, diverse building entrances, OLT or edge locations, upstream circuit rates, backup power, spare stock, crew depth, restoration target or failure-time usable capacity. The proper operating grade is Medium; the physical-resilience grade remains Weak.

The handhole is where two contracts can become one route

At 06:10, before the traffic thickens on an El Vigía commercial street, a crew lifts the lid of a fibre handhole. Two colour-coded conduits enter from opposite sides. On a sales order they could be Circuit A and Circuit B. On a router they could support two sessions. Yet inside the chamber they may turn into one larger duct beneath the next junction, pass through the same flooded closure, enter the same building sleeve or depend on the same powered cabinet.

This is an analytical inspection scene, not a report of a Fibertech handhole, circuit or outage. No public source reviewed for this article identifies such a chamber as company infrastructure. The scene matters because it describes the point at which apparently separate services either become genuine physical diversity or collapse into one failure domain.

The distinction is recognised in the engineering literature. ITU-T L.250 separates fibres in one cable, different cables on the same route, different cables on different routes from one central office, and different routes from different central offices. Those options do not provide equal protection. Two cables on the same route may survive a fault in one cable, but not necessarily excavation, flooding, road collapse or another event that breaks the shared route. Different routes from different offices address a wider set of common failures, including loss of one office's power.

Fibertech's public evidence does not disclose where it sits on that scale. Its current website says the company uses fibre, offers service to homes and businesses, monitors continuously and is expanding its network. It does not say whether the access plant is owned, leased or mixed; whether feeders form a ring or terminate as branches; whether any two external circuits use separate carriers; or whether they enter a network site through opposite sides of a building.

Those omissions are not proof of bad design. Small operators rarely publish security-sensitive route drawings. But they prevent a stronger resilience claim. A provider can have two invoices, two optics and two BGP sessions while still exposing both to the same backhoe. It can have a fibre ring whose two arms meet in one bridge duct. It can buy two services from the same upstream through the same local carrier. It can use different streets but one unprotected edge router. Each design looks redundant at one layer and concentrated at another.

The opening question is therefore not, "Does Fibertech have fibre?" It plainly markets fibre. Nor is it, "Does Fibertech have an ASN?" It does. The question is whether the path from a subscriber's optical network terminal to the public internet contains a genuinely independent route at every mandatory transition: drop, splitter, feeder, OLT, aggregation switch, edge router, upstream transport, handoff facility and power supply.

The handhole is the right place to begin because it forces commercial language to meet civil infrastructure. A second circuit only changes the outage if it survives the same event. Until a sanitised route-diversity statement, entrance drawing or tested failover record establishes that separation, the prudent conclusion is not that Fibertech has no alternative. It is that the public cannot tell whether a second logical option becomes a second physical route before both disappear under the same street.

A live operator, more precisely bounded than its sales page

Fibertech's operating hypothesis can be upgraded from "thin and uncertain" to "present, but incompletely documented." The evidence is cumulative.

The company website names Corporación Fibertech C.A., gives an address at Sector La Inmaculada, Calle 10 with Avenida 11, local 11-04, El Vigía, and publishes separate telephone channels for customer attention and technical support. It offers unlimited fibre packages at 450 Mbps for $25, 650 Mbps for $30 and 1 Gbps for $60. It says dual-stack IPv4 and IPv6 are available and describes commercial uses including point-of-sale terminals, cameras, internal systems and cloud services.

The site links to a live customer payment portal. The portal asks for an identification number and presents itself as "Fiber King", rather than repeating the Fibertech legal name. That naming difference should not be promoted into a separate company or network. The evidentiary point is narrower: Fibertech's own site directs customers to an operational account-management surface. It supports the existence of a billing relationship, but it does not reveal how many accounts are active, who owns the portal software, where customer data is processed or whether the portal remains usable during a local access failure.

The public routing edge is also active. LACNIC's AS273994 record assigns the autonomous system to CORPORACION FIBERTECH C.A. and places the registrant at the same El Vigía street intersection. LACNIC's IPv6 record assigns 2803:58d0::/32 to the same legal name and address. At the publication timestamp, both IPv4 and IPv6 were visible in the global routing system.

A secondary IPinfo profile describes AS273994 as a consumer ISP, reports a pronounced day-and-night activity pattern and geolocates sampled router addresses to El Vigía. Those observations are consistent with an active access network. They are not a subscriber count, a verified router map or proof that every address is physically in the city. Third-party geolocation and behavioural classification remain inferences.

Together, the website, support contacts, payment surface, registry identity and current routes justify a Medium normal-operation grade. None individually would be enough. A website can outlive a network; an ASN can remain registered while unused; a route can serve infrastructure rather than retail customers; a payment page can be a generic tool. The combination is materially stronger.

The grade stops at Medium because important operational denominators remain absent. There is no public customer count, premises-passed figure, active ONU count, service-area polygon, regulator row under the exact legal name, traffic graph, uptime series or attributable incident history. The operator is visible. Its scale and continuity design are not.

The legal company, the Fibertech brand, AS273994 and the physical plant are four boundaries

Resilience analysis fails quickly when corporate, commercial, routing and physical identities are treated as interchangeable.

The legal subject is CORPORACION FIBERTECH C.A. That exact name appears in LACNIC's ASN and IPv6 records and on the public website, with small differences in capitalisation and punctuation. The public brand is Fibertech. The payment surface uses Fiber King. AS273994 is a routing identity registered to the legal company. None of those labels, by itself, identifies who owns a duct, fibre cable, OLT, street cabinet, edge router, vehicle, generator or upstream transport circuit.

The IPv6 boundary is comparatively clear. The /32 is registered by LACNIC directly to the company. Even there, "holds an IPv6 allocation" is not the same as "uses every address", "assigns IPv6 to every subscriber" or "has a physically independent IPv6 path." IPv4 and IPv6 can share the same fibre, router, upstream and power system.

The current IPv4 boundary is different. The RIPE Database record for 178.19.41.0/24 names FIBERTECH_NET001, country VE and organisation ORG-CFC5-RIPE, but marks the block ASSIGNED PA and lists mnt-ipbroker as maintainer. RIPE's entity-status documentation explains that ASSIGNED PA space is assigned by a member from its allocation to an end user. The associated organisation record names Corporacion Fibertech, C.A. and was created in May 2026.

That supports a precise statement: Fibertech currently originates a provider-aggregatable IPv4 assignment associated with its legal identity. It does not support saying that the company owns the parent allocation, that IPBroker carries the physical circuit, or that the /24 identifies a particular site. Address administration, transit and fibre transport can involve different parties.

The historical IPv4 block illustrates the same problem. AS273994 previously originated 154.47.36.0/22. The current ARIN parent record places the encompassing address space with Cogent Communications. That does not prove Fibertech bought direct Cogent transit, used a Cogent-owned fibre into El Vigía or housed equipment in a Cogent facility. A provider-address assignment can be delivered through an intermediary, announced through another upstream and later returned or reassigned.

The physical plant boundary is the least visible. Fibertech's website says "fibre optic", but it does not identify whether the company owns the glass from each customer to its aggregation point. A regional provider can own drops and splitters while leasing feeder fibre. It can own an OLT but buy wavelength or Ethernet transport. It can place equipment in a third-party room. It can outsource civil work while retaining service responsibility. All are plausible; none is established here.

This separation also governs responsibility during a failure. The subscriber calls Fibertech. Fibertech may diagnose a loss. Yet a property owner may control building access, a utility may control a pole, a civil contractor may control a street opening, a fibre lessor may control a damaged span, and an upstream may control the far-end port. The customer experiences one service, while restoration can cross several contracts and safety boundaries.

The article therefore attributes only what the evidence supports: the company operates the brand, holds AS273994, holds the IPv6 /32, is associated with the current IPv4 /24, and markets local fibre service. It does not attribute TotalCom's network, Cogent's parent address space, IPBroker's allocation, DigitalOcean's portal hosting or any unverified street asset to Fibertech.

A location check is not a coverage map

The most defensible physical anchor is the public address in Sector La Inmaculada. An OpenStreetMap-derived Photon result for La Inmaculada corroborates the district within El Vigía and the municipality of Alberto Adriani. It does not return a verified Fibertech building point at local 11-04, and it certainly does not return a network route.

The company's own coverage language is deliberately address-specific. Prospective customers are asked to send their exact location so staff can verify availability. That is useful commercial honesty: a fibre network is not uniformly present across a city merely because a provider names the city. It also means the public site supplies no coverage polygon, street list, neighbourhood inventory or premises-passed total.

The absence matters because failure impact follows topology, not municipal labels. Ten streets served from one feeder create a different risk from ten streets split across protected feeder paths. A dense block with several splitters may expose many users to one closure. A long branch toward a peripheral neighbourhood may expose fewer users but require more travel, access and cable. Without feeder areas, splitter domains, OLT associations or route lengths, even a rough blast radius would be fabricated.

Regulatory geography remains unresolved. A publicly reproduced CONATEL list dated 4 December 2025 contains authorised internet-service providers, their tax identifiers, habilitation numbers and coverage areas. A search of that document did not find the exact legal name CORPORACION FIBERTECH C.A. This is a material evidentiary gap, not a finding of unauthorised operation. The reproduction could omit an amendment, the authorisation could post-date the document, a different legal name could be used, or the public copy could be incomplete.

The correct map is therefore sparse:

1. A verified company and registry address in El Vigía.
2. First-party marketing to homes and businesses, with availability checked by exact location.
3. Third-party network geolocation broadly consistent with El Vigía.
4. No public route, ring, coverage polygon, POP, cabinet, OLT, upstream handoff or building entrance.

Connecting those points with invented lines would produce a map that looks informative while erasing the central fact: the physical continuity surface is not published.

The nearly 87-day IPv4 gap needs restraint

Fibertech's route history is the most specific unusual feature in the public record. It is also the easiest one to overstate.

RIPEstat's one-year announced-prefix history shows the historical IPv4 block 154.47.36.0/22 visible from 30 July 2025 until 23 March 2026. It shows the current 178.19.41.0/24 beginning on 18 June 2026. Between those observations lies almost 87 days in which RIPEstat did not record an IPv4 prefix originated by AS273994.

That interval is not evidence of an 87-day customer outage.

Fibertech may have delivered IPv4 through address space originated by another ASN. It may have used carrier-grade NAT behind an upstream assignment. It may have operated a smaller or collector-invisible announcement. Customers may have retained IPv6 while IPv4 arrangements changed. The website and portal do not provide a dated service log that resolves these possibilities. Public route collectors show what was announced under AS273994, not every way a customer might have reached the IPv4 internet.

The route change still matters. At the historical snapshot on 22 March 2026, the BGP state for 154.47.36.0/22 contained 381 collected paths, all with AS264778 immediately before AS273994. At 08:00 UTC on 16 July, the current /24 BGP state contained 379 paths, again all with AS264778 immediately before AS273994.

The address supplier changed; the visible immediate upstream did not.

The numerical footprint also changed from 1,024 IPv4 addresses in a /22 to 256 in a /24. That is a fourfold reduction in addresses originated under the company's ASN. It is not a fourfold reduction in customers, bandwidth or service capacity. NAT allows many customer devices to share public addresses. A provider can deliver a 1 Gbps retail line while using no dedicated public IPv4 address for that household. Conversely, a large public block can sit mostly unused.

RPKI adds another bounded distinction. The current 178.19.41.0/24 validation result is valid for origin AS273994 with maximum length /24. The current IPv6 validation result is valid with maximum length /48. The historical 154.47.36.0/22 result is now unknown, with no validating ROA in the current result.

As the RPKI community's explanation of route-origin validation makes clear, a valid result authorises an ASN to originate a prefix. It does not authenticate the AS path, measure performance or keep a cable powered. The current valid ROAs improve origin security. They do not explain the physical route or the IPv4 transition.

The useful conclusion is narrower than an outage story. Fibertech has changed the contractual and administrative wrapper around its public IPv4 service while continuing to appear behind the same visible immediate upstream. That shows operational change. It does not show whether the underlying transport changed street, carrier, facility, router or power domain. An address migration can happen entirely above the handhole.

One observed neighbour is a warning label, not a fibre diagram

At 08:00 UTC on 16 July, Fibertech's RIPEstat routing-status snapshot showed one IPv4 prefix covering 256 addresses and one IPv6 /32, equivalent in routing-space arithmetic to 65,536 /48 networks. The IPv4 route was visible to 325 of 326 responding IPv4 RIS peers; IPv6 was visible to 319 of 321 responding IPv6 peers. The snapshot counted one observed neighbour.

The AS-neighbour dataset identifies that neighbour as AS264778, TotalCom Venezuela C.A. The IPv6 BGP-state snapshot contained 365 paths, all ending with AS264778 immediately before AS273994. The IPv4 result showed the same immediate predecessor.

This is strong evidence of one visible logical boundary in normal operation. It does not establish one physical circuit.

RIPE RIS gathers BGP information from route collectors and peers. Its collector documentation explains that collection includes both local and multi-hop sessions. The resulting AS paths describe routing information observed from many vantage points. They do not expose the number of Ethernet services between two companies, the cable carrier, street route, handoff room, router pair or power feeds.

Several physical designs remain compatible with the same public evidence:

• One Fibertech edge router could have one circuit to one TotalCom router.
• Two Fibertech routers could have two circuits to the same TotalCom ASN.
• Two circuits could use different fibres in one cable.
• They could use separate cables in one duct.
• They could take different street routes but enter one building sleeve.
• A second upstream could exist in standby and remain invisible until failover.
• A backup could use provider-assigned addresses rather than originate the same prefixes.

Public BGP cannot choose among these designs.

The same caution cuts in the other direction. One neighbour should not be dismissed as meaningless. If all normal and backup paths for both prefixes ultimately depend on AS264778, then a commercial, policy, edge or common-facility failure at that boundary can affect both address families. If no other provider is authorised and physically ready to carry the routes, upstream independence does not exist merely because there are two circuits.

The static snapshot also cannot show failure behaviour. A dormant route may appear only after withdrawal. A route may converge but lack enough bandwidth for the surviving load. A second circuit may carry control traffic but not all prefixes. IPv4 may fail over while IPv6 does not, or vice versa. The only persuasive evidence would be a dated failover test that records which prefixes moved, how quickly they converged, what traffic the surviving path carried, and whether the test removed a genuinely independent physical element.

For now, "one visible upstream" is best treated as a warning label. It identifies the boundary that deserves the next question. It does not answer whether the boundary consists of one cable, two cables, one site or two sites.

TotalCom's wider network stops at Fibertech's demarcation

TotalCom is itself a live and more connected network. LACNIC registers AS264778 to TotalCom Venezuela C.A. in Maracaibo. At the same publication timestamp, RIPEstat reported 12 IPv4 prefixes, six IPv6 prefixes, broad collector visibility and ten observed neighbours. That wider topology can improve TotalCom's own ability to reach the internet.

PeeringDB gives a more tangible but still bounded fact. TotalCom's operator-maintained profile lists one operational 100 Gbps interface at NAPVE MCBO in Maracaibo, with route-server participation. The exchange profile identifies NAPVE MCBO as a Maracaibo exchange. NAP VE's own about page says it operates exchange points in Caracas, Maracaibo and Valencia.

None of that 100 Gbps belongs to Fibertech merely because TotalCom is the visible adjacent ASN. It is not evidence of Fibertech's circuit rate, transit commit, burst entitlement, traffic level, reserved headroom or physical route to Maracaibo. The BGP adjacency could be delivered through a different city, a remote service, another carrier or a private handoff not identified in PeeringDB.

This is the demarcation principle: redundancy beyond TotalCom helps Fibertech only if the Fibertech-to-TotalCom path survives. A well-connected upstream cannot repair a cut before the handoff. A 100 Gbps exchange port cannot power Fibertech's edge router. Multiple TotalCom neighbours cannot create a second Fibertech building entrance.

The reverse is also true. Fibertech could have two physically diverse links to TotalCom even though the public sees one ASN. If so, the company could disclose the diversity in a sanitised form without naming exact routes: separate carriers, separate street approaches, separate buildings or meet-me rooms, independent edge devices and tested failover. No such disclosure was found.

TotalCom's public scale therefore provides upstream context, not borrowed resilience. Fibertech's continuity must be demonstrated on Fibertech's side of the boundary.

Three meanings of one gigabit

The 1 Gbps plan is the largest number on Fibertech's sales page. It is also the number most likely to be mistaken for network capacity.

First, it is a marketed access rate. The website describes a 1 Gbps unlimited fibre plan for $60 per month, alongside 450 Mbps and 650 Mbps plans. It does not publish whether the rate is symmetrical, the measurement method, minimum throughput, contention policy, PON standard, split ratio or busy-hour performance. A plan rate is the maximum service proposition at one subscriber interface, not a promise that every user can draw that rate simultaneously.

Second, there is installed capacity. That would include OLT line cards and ports, optical split, feeder fibres, aggregation ports, edge-router interfaces, upstream circuits, power systems and cooling. No attributable installed total was found. Brand logos for networking vendors on the website do not prove particular models, quantities or architecture.

Third, there is failure-time usable capacity. This is what remains when one mandatory component is removed. It cannot be calculated by adding retail rates, IP addresses or an upstream's exchange ports. End-to-end usable service is constrained by the smallest surviving element:

surviving access path -> powered aggregation -> live edge -> surviving transport -> accepted route -> remaining upstream headroom

If any mandatory element has no alternative, the usable capacity for the dependent users can become zero even while other parts of the network remain healthy. If an OLT loses power, spare transit bandwidth does not help its subscribers. If one common duct carries both external circuits and is cut, two upstream orders do not help. If a second path survives physically but has no reserved bandwidth, routes may recover while applications become congested.

The address estate supplies no bandwidth answer. A /24 contains 256 IPv4 addresses. An IPv6 /32 can be divided into many customer prefixes. Neither number measures bits per second. The current IPv4 block is smaller than the historical /22, yet Fibertech's retail page now markets up to 1 Gbps. That coexistence is not contradictory because addressing and transport are different capacity classes.

TotalCom's 100 Gbps NAPVE interface is equally non-additive. It is one declared interface in the upstream network. Fibertech's share may be a fraction, may not traverse that port, and may be limited elsewhere. Treating it as Fibertech capacity would be like treating the width of a national motorway as the capacity of an undisclosed driveway.

Sold capacity is also unknown. The website does not publish subscriber totals or plan mix. IPinfo's consumer-network pattern suggests household activity but cannot count homes. The payment portal cannot count customers from its public page. Without sold load, current utilisation and protected reserve, no credible N-1 throughput can be stated.

The honest capacity statement is therefore qualitative. Fibertech markets gigabit-class access and operates broadly visible routes. The design, installed, lit, powered, sold, reserved and failure-usable quantities behind that offer are undisclosed.

What remains usable when something fails

A resilience assessment becomes useful only when it names the failed component and follows the surviving service to the user.

A customer-premises power failure can disconnect one home or shop even if Fibertech's network is healthy. The ONU, Wi-Fi router and any local switch need electricity. A battery-backed router without a battery-backed ONU still loses optical service. A customer UPS also cannot help if the upstream splitter, OLT or edge site is dark.

A drop-fibre failure may affect one premises. The relevant capacity for that customer is zero until the drop is repaired or temporarily bypassed. Other users on the same splitter may remain online.

A splitter or distribution-cable failure can affect an unknown group of premises. The group size depends on the optical design and actual occupancy, neither of which is public. A nominal 1 Gbps plan does not change the shared physical dependency.

A feeder cut can remove one or more splitter trees. If the feeder is part of a protected ring with a functioning opposite arm, traffic may be switched. If both arms share the same handhole, bridge, duct or entrance, the ring can fail as one route. Public evidence does not identify Fibertech's design.

An OLT or aggregation failure can affect every active subscriber dependent on that equipment. A spare card helps only if it is compatible, configured and available. A spare chassis helps only if technicians can reach the site, power it and reconnect fibres. No spare inventory or site design is public.

An edge-router or site-power failure can remove both current prefixes while leaving customer optical light levels apparently normal. Subscribers may see a lit PON indicator but no public reachability. A redundant router in the same rack and on the same power feed addresses equipment failure but not rack or power failure.

Loss of the TotalCom boundary is the clearest logical external test. Both current prefixes were observed through AS264778. If that adjacency disappears and no hidden or standby alternative activates, public IPv4 and IPv6 reachability from AS273994 would be lost. This is a conditional result, not a claim that Fibertech has no backup.

A wider TotalCom failure may or may not reach Fibertech. TotalCom has multiple observed neighbours, but the effect depends on routing policy, traffic engineering and capacity. A path can exist in BGP yet be too congested to deliver the sold load.

A simultaneous power and street-access failure is more demanding. Rain can flood a chamber, damage a closure, block a road and interrupt electricity. The network may need a civil crew, utility clearance, pumping, fibre technicians, replacement cable and upstream coordination in sequence. The slowest dependency determines restoration.

Failure-time usable capacity is therefore not one number waiting to be discovered. It is a matrix:

Failure	Users potentially affected	Capacity that can be claimed publicly
One customer power supply or drop	One premises	Unknown for that premises; potentially zero until local restoration
One splitter or distribution branch	Premises downstream of that branch	Unknown because split and occupancy are unpublished
One feeder or common duct	All downstream branches sharing it	Unknown; protected-path existence and separation are unpublished
One OLT, aggregation switch or local site	All customers dependent on that device or site	Unknown; equipment associations and backup power are unpublished
One external circuit	Customers whose traffic cannot move elsewhere	Unknown; circuit count, routing policy and spare bandwidth are unpublished
Entire visible AS264778 boundary	Potentially both originated address families if no alternative activates	No tested surviving rate is public
One field crew or missing spare	Any unresolved fault requiring that skill or part	Restoration throughput and delay are unknown

The table deliberately avoids a customer count. Without a node-to-premises dependency map, any number would be invented. It also avoids declaring zero for the entire network. A local cut can leave other areas operating; an external routing failure can leave local management reachable; an externally hosted portal can remain online from other networks.

The website itself illustrates that last distinction. Public DNS places fibertech.com.ve behind Cloudflare rather than AS273994, while the payment portal's address is registered to DigitalOcean. Those external surfaces may remain reachable during a Fibertech routing outage. They can help customers obtain information or make payments, but only if staff can access them and affected users have another connection. An online website is not evidence that the access network is online.

Recovery begins with diagnosis, not a speed test

When fibre service fails, the first job is to locate the layer. A speed test is useful only after packets can travel. Before that, support must distinguish customer power, Wi-Fi, ONU, optical loss, feeder, active equipment, routing and upstream failure.

Fibertech's website separates a customer-attention number from a technical-support number and claims 24-hour monitoring and support. That is evidence of an intake and monitoring surface. It does not establish a staffed field shift, response time, simultaneous-incident capacity or authority to enter every site.

The recovery sequence is well understood in principle. ITU-T L.25 describes post-fault work that includes receiving an alarm or customer report, distinguishing fibre-line failure from transmission-equipment failure, testing to locate the fault, dispatching technicians, repairing the cable, verifying the repair and returning to normal operation. It also describes rerouting and temporary restoration cable where those options exist.

ITU-T L.93 explains why installation and maintenance records matter. OTDR testing can measure connection loss and reflectance, and a new trace can be compared with a reference trace to identify where the optical condition changed. The independent Fiber Optic Association restoration guide makes the operational point plainly: restoration is faster when the cable plant was documented during installation, suitable test equipment is ready, and spare components and procedures were prepared before the emergency.

Underground plant adds water and access. ITU-T L.315 notes that water entering underground closures or cabinets can increase the risk of significant degradation to fibres and connectors. In a flooded handhole, the task may include making the chamber safe, removing water, identifying the correct closure, checking cable and splice condition, and preventing contamination before a splice is remade.

For Fibertech, every stage after remote diagnosis is publicly unmeasured.

There is no disclosed OTDR baseline, fibre-management system, depot, crew origin, vehicle count, safe-entry arrangement, contractor roster or spare inventory. There is no public list of compatible ONUs, optics, OLT cards, splitters, closures, feeder cable, drop cable, batteries or edge-router components held in El Vigía. There is no stated escalation boundary for leased fibre or TotalCom transport.

Local support remains valuable. A nearby technician who knows the street plant can shorten diagnosis and travel. A customer-facing office and separate support channel reduce the distance between complaint and acknowledgement. But "local" does not answer whether the correct specialist is available at 03:00, whether two faults can be handled at once, whether the chamber can legally be opened, or whether the required cable and closure are in stock.

Recovery responsibility is especially important when assets are mixed. If Fibertech owns the customer drop but leases the feeder, it may repair one and wait for another party on the other. If it owns the feeder but uses a third-party duct, street access may require permission. If TotalCom supplies transport to the edge, Fibertech can test its local interface while the upstream controls the far end. A clear responsibility matrix can be more valuable than a generic "24/7" statement.

The meaningful public metrics would be operational rather than promotional: median time to acknowledge, isolate and restore by fault class; percentage of incidents resolved remotely; number of concurrent fibre crews; spare availability by critical component; battery runtime under tested load; and dates of successful failover exercises. None was found.

The affected-user number is the missing denominator

Fibertech's own sales language identifies the kinds of users exposed to failure. Homes rely on the connection for work, study, streaming and gaming. Businesses may depend on point-of-sale systems, cameras, internal administration, customer service and cloud tools. Those are service use cases, not a customer list.

The number of affected users cannot be recovered from the public address blocks. A /24 is not 256 customers. The IPv6 /32 is not 65,536 customers. One payment portal is not one account per page view. IPinfo's day-and-night signature is not a household census.

The impact denominator must come from physical dependencies:

• active ONUs behind each splitter;
• splitters behind each distribution cable;
• distribution branches behind each feeder;
• feeders behind each OLT;
• OLTs and switches behind each power site;
• sites behind each edge and external circuit;
• customers whose traffic can or cannot move to an alternate path.

None of those associations is public.

That means the most responsible affected-user statement is conditional. A damaged customer drop may affect one premises. A flooded handhole containing a shared distribution cable may affect all downstream premises on that cable. A common feeder, OLT, edge site or external handoff may affect a larger but unknown set. A loss of the only functioning public upstream path could affect customers across the service footprint, but a hidden backup may limit the impact.

Business impact can also exceed the headcount. One shop may represent one subscription but lose card transactions, remote stock access and cameras. One home connection may support several people working or studying. Conversely, a business may have mobile backup or a second provider. The article cannot assign either vulnerability or redundancy to unnamed customers.

This missing denominator is not a reason to avoid impact analysis. It is a reason to tie every claim to the failed component and refuse false precision. "Potentially all users downstream of the common point" is more useful than an invented citywide number.

Rain, electricity and access can turn one cut into a chain

El Vigía's hazard context makes the handhole more than a metaphor, but it must remain separate from company incident evidence.

In February 2024, a report citing Mérida civil-protection officials said heavy rain affected Alberto Adriani and other municipalities, damaged homes, increased river levels and led to preventive road closures including the Onia route connected with El Vigía. The report does not identify Fibertech infrastructure or customers.

In June 2025, an IFRC-linked situation report recorded heavy rainfall, river increases, road damage, fallen electricity poles and power interruptions across parts of Mérida. Again, it does not document a Fibertech outage or an event in the company's exact service area.

The relevance is mechanical. Heavy rain can raise water in chambers, obstruct roads and delay safe entry. Power interruption can affect customer equipment and active network sites at the same time. Damaged roads can prevent a crew or replacement cable from reaching a fault. A utility pole failure can combine electricity and telecom damage where cables share structures.

Resilience against this chain requires more than an alternate BGP path. It requires separated routes, sealed and maintained closures, safe access, tested backup power, documented fibre, spare materials, trained labour and clear authority to act. Public evidence does not show whether Fibertech has those protections.

The correct conclusion is therefore exposure, not attribution. El Vigía and Mérida have experienced conditions capable of stressing underground, aerial, power and transport systems. No reviewed evidence says those conditions caused a Fibertech failure.

What proof would change the grade

Fibertech could materially improve confidence without publishing exact cable coordinates or commercially sensitive traffic.

First, it could publish a sanitised service and topology statement: municipalities or neighbourhood groups actually served; whether access is wholly fibre; whether core feeders are ringed; and whether the ring arms use different street routes. A diagram can show failure domains without showing handhole coordinates.

Second, it could disclose the ownership boundary: which drops, splitters, feeders, ducts, poles, OLTs and edge devices are owned; which are leased; and which party controls restoration. The name of every landlord is unnecessary. The division of responsibility is enough.

Third, it could describe entrance and site diversity: whether external circuits enter from different streets, terminate in separate rooms or racks, use separate routers, and depend on independent power. Two services in one conduit should be identified as cable diversity, not route diversity.

Fourth, it could publish an upstream resilience statement: number of active external circuits; number of upstream ASNs able to carry customer traffic; whether both IPv4 and IPv6 can move; and the date of the last failover test. A test should remove a physical circuit or edge device, not merely administratively shut one BGP session.

Fifth, it could disclose failure-time capacity in ranges: normal busy-hour utilisation, largest-circuit loss, surviving committed capacity and whether traffic is prioritised during degradation. No customer traffic graph is required. The key question is whether the surviving path can carry the essential load.

Sixth, it could publish power and field readiness: minimum tested battery runtime by site class, generator coverage, refuelling arrangements, number of on-call fibre crews, critical spare categories and restoration targets by fault class.

Seventh, it could clarify regulatory identity by publishing the applicable CONATEL authorisation, tax identifier and authorised geography under the exact legal entity. That would close the current gap between a clear LACNIC identity and the absence of an exact-name match in the dated public list reproduction.

Each item would answer a different question. None should be collapsed into a generic "redundant network" badge. A provider can be strong in routing security and weak in power, strong in local repair and weak in upstream independence, or strong in route diversity and weak in spare capacity.

Fibertech is operating; its continuity architecture is not public

Fibertech no longer looks like a dormant name. Its current retail offer, local contact details, payment surface, LACNIC records, current IPv4 assignment, valid route-origin authorisations and broadly visible dual-stack routes form a coherent operating picture. The normal-state grade is Medium rather than Weak.

The resilience conclusion is less generous because the strongest facts are logical. AS273994 currently originates a /24 and an IPv6 /32; both appear behind TotalCom. Its historical IPv4 /22 also appeared behind TotalCom before the address estate changed. These facts identify a stable visible upstream boundary across an address transition. They do not identify a stable physical route.

Under the street, the decisive evidence is missing: ownership, duct separation, ring closure, building entrances, edge and power separation, spare bandwidth, repair responsibility and tested recovery. A second circuit may exist. A protected ring may exist. A field team may be ready. The public record cannot promote those possibilities into facts.

The final grade is therefore deliberately split: Medium for current operation and routing identity; Weak for physical continuity and failure-time usable capacity. The fastest way to raise it is not another speed claim. It is evidence that the two colours entering the handhole remain two independent routes after they leave it.