Observed incident records

Every incident below was detected by RELIASTRA's own measurement: consecutive failed probes against one endpoint from a named region, confirmed by the published detection rule before the record exists. A record states what was observed, from where, and when - it does not infer a vendor-wide outage, and it is never written from a status page.

Search

Active: vendor mistral. Clear all filters to return to the full record.

Incident records - 1

Newest first by detection start, printed in UTC. 'Failures' is the count of consecutive failed observations the rule counted; the record itself holds the window and the evidence references.

26 Sep 2026 17:10:56 UTC Mistral AIOfficial status pageus-eastOpen

HTTP 4xx responses2 failuresopen at last readRecord

How to read a record

An incident here means one thing: RELIASTRA probes in the named region recorded consecutive failures against one endpoint for this dependency, across the stated window. The two-failure confirmation threshold and two-success recovery rule are the whole gate - no timing heuristics, no scoring, no inference.

A record never claims the vendor itself was down. One endpoint failing can be a regional probe path, a rate limit, or a genuine service failure - the record states what was measured and stops. Where evidence-grade detail was published for the same window, the record page holds the full provenance.