Monthly reliability, including the failures.

How the numbers are calculated

• • Availability is the share of recorded canary checks that passed, not time-weighted uptime. Missing checks are not assumed successful; unconfigured optional integrations are excluded.
• • Monitoring coverage counts completed 15-minute UTC intervals containing a valid platform snapshot, whether passing or failing. The current unfinished interval is excluded. Missing intervals are unknown, not outages or successes; with a truncated sample, older observations may exist outside this report. A snapshot does not prove every service was checked or that the whole interval was healthy. At most the first 100 gaps are listed.
• • Execution aggregates use at most the latest 1,000 logged signals; check aggregates use at most the latest 1,000 snapshots. Incidents use at most 200 rows. Coverage reports sampled versus total stored rows; a partial sample is not a complete monthly history.
• • Legacy Desktop history uploads without verified broker receipts are excluded from execution proof. A recorded execution status is receiver-reported evidence, not independent broker certification.
• • Execution success rate uses terminal attempted signals only: executed divided by executed plus failed.
• • Latency is reported by recorded source, parse, dispatch, broker-acceptance and confirmed-fill stages. Missing stages remain missing and are never inferred.
• • Broker rejection causes are grouped into public categories; private messages, account identifiers and credentials are never published.

Previous calendar month (2026-08), bounded sample: checks passed 88.462%; execution success 33.33%.