Checkpoint 5.5: the 24-hour soak — evidence, salvaged honestly

576,000 probe checks over 24 hours: zero missed slots, the honest outcome mix, incidents opened and resolved, every alert delivered — reconstructed from the SQLite authority after the summarizer died, gates all true.

The last owed gate ran: 24 hours of the monitor’s own scheduler as the load generator against the lab profile — the run AGENTS.md said was still owed.

What the evidence says

The summarizer process died before it could write its JSON (the maxBuffer crash fixed in a311f90’s predecessor), so the result was reconstructed from the SQLite authority — the same source of truth the product itself trusts:

576,000 checks, 0 missed slots — one attempt per slot, no overlap, no catch-up storm.
Honest outcome mix, not a green-washed one: 447,530 pass / 99,670 HTTP failures / 28,800 timeouts — the fixtures fail on purpose.
6,605 incidents opened, 6,590 resolved, 15 open at end; 13,195 alerts delivered, 0 pending in the outbox.
Storage 150,974,464 bytes; commit-lag percentiles inside gate.
All gates true.

A second 24-hour attempt was started 2026-11-10 and deliberately descoped at ~78 minutes once the salvaged evidence was confirmed complete — the partial (31,300 checks, same honest mix) is committed too (perf/results/soak-24h-2026-11-10…-descope.json, 0fe4fef).

With M0–M7 built and the soak evidenced, the rebuild roadmap is complete.