Summary

  • The heat map measures added latency against the fastest bucket in the visible window. Start after a sustained reroute and that reroute can become the new green baseline.
  • Direction merging is on by default and keeps the lower RTT in each bucket; packet loss has its own disrupted signal.

Your outage is green. The chart is doing exactly what you asked; the question is whether you asked the right time window.

RIPE Atlas introduced Anchor Mesh Status on 20 October 2026. The feature turns the existing ping and traceroute measurements between Atlas anchors into a heat map of latency and packet loss. It does not schedule new measurements. Each anchor already measures the others; the new part is the view that lets a reader pick corridors, time ranges and address families, then follow a coloured square into path diagnostics.

Here is the small calculation with the large consequence. In each time bucket, the chart averages the median round-trip times reported for that bucket. For the visible range, it chooses the lowest bucket as the corridor’s baseline. Every colour then shows the difference from that best bucket in milliseconds. If the path rerouted on Monday and your selection begins Wednesday, Wednesday’s slower journey can be the fastest thing in view. The chart subtracts that Monday-sized detour from itself and paints the result green. A long incident can look like a settled routine because the window forgot what “before” looked like.

RIPE’s own guide says to include time before the event. Its Baltic Sea example makes the point: the documented Lithuania–Sweden anchor corridors move from about 17 ms to about 37 ms after a cable cut in November 2024, stay elevated for roughly eleven days and return after repair. A window beginning before the cut shows the change; a window beginning during the detour asks a different question.

A second switch changes what “the path” means. Direction merging is on by default. For each bucket, the display keeps the lower RTT of A→B and B→A and presents that best case in one row. That is a tidy overview, and it can hide the slower direction. Open the pair, use Reverse, and compare the unmerged directions when the service depends on requests and replies travelling well both ways.

The palette also separates delay from silence. More than 20% packet loss fades a cell. At 100% loss, there is no RTT to measure, so the bucket appears as disrupted rather than as inflated latency. Grey means there were no samples. A fast-looking number that violates the physical-distance check gets its own warning and is excluded from the baseline. The summary strip keeps “latency inflated” and “disrupted” as separate counts.

Sources: RIPE NCC’s October 2026 Member Update; RIPE Atlas Anchor Mesh Status documentation; the live Anchor Mesh Status view; the 20 October Atlas release notes; the RIPE Atlas anchoring-mesh ping measurement API.