Checkpoint 3.2: a status page that outlives the monitor

status.json rewritten every beat — atomic, allowlisted fields only, self-expiring — rendered by a static page with zero runtime dependency on the monitor.

at a glance

Goal: A public status artifact that keeps answering after the monitor dies — sanitized, self-expiring, hostable anywhere.
Built: status.json rewritten every SMON_STATUS_INTERVAL_S (atomic tmp+rename): sha256 service alias, status, incident_open_since only; TTL declared in-file; apps/status/index.html static renderer.
Result: allowlist/TTL unit tests + writer-beats-file test · the alias bug found live: UUIDv7 prefixes collide — aliases now sha256(id)[:12]
Code: akira-shirase-oss/monitoring-service · main @ 570259a

What changed: the server rewrites status.json every beat into SMON_STATUS_DIR — atomic tmp+rename, so a reader never sees a torn file. Allowlisted fields only: a hashed service alias, the status, the incident-open time — nothing a management viewer would see and a public viewer shouldn’t. apps/status/index.html renders it with zero runtime dependency: copy the two files to any host and the status page outlives the monitor itself.

The alias bug, found live: my first alias was an id prefix — and two monitors created milliseconds apart produced the same alias, because UUIDv7 leads with its timestamp. Aliases are sha256(id)[:12] now: stable, opaque, uniformly spread.

Proof: allowlist and TTL unit tests plus the writer-beats-file test — the snapshot’s schema is the contract, pinned by test.