fix: add /readyz Fly check so a dead DB fails health

[BeLazy167]

Problem

/healthz returns a hardcoded {"status":"ok"} and never touches the store (internal/api/server.go:302). It was the only check wired into [checks].

On 2026-09-03 argus-db filled its 10 GB volume. Postgres then crash-looped roughly every 2 seconds:

PANIC: could not write to file "pg_logical/replorigin_checkpoint.tmp": No space left on device
LOG: checkpointer process was terminated by signal 6: Aborted
LOG: all server processes terminated; reinitializing

This ran for three days. Fly reported the app healthy throughout, because /healthz cannot observe the database. The failure surfaced only as "the site loads but nothing works" — the static Next.js shell served 200 while every data fetch failed.

/readyz already existed and pings the pool (server.go:308). It correctly returned 503 for the whole outage. Nothing was watching it.

Change

Adds a second Fly check on /readyz. health keeps liveness semantics on /healthz; ready answers whether the machine can actually serve.

ready uses a 20s grace_period because the pool opens after the listener on boot, and a 30s interval because each probe costs a round trip to Postgres.

Verification

• fly config validate → Configuration is valid
• go build ./... and go vet ./... pass
• /readyz is registered unauthenticated next to /healthz (server.go:133), so the Fly checker reaches it
• Confirmed against the live outage: /healthz 200 and /readyz 503 while Postgres was down; both 200 after the volume was extended

No Go code changes.

🤖 Generated with Claude Code

https://claude.ai/code/session_01GJehDM4aXz1AkJrTaX2mSL

[BeLazy167]

Wrong repo — reopened as BeLazy167/argus-private#284, where development PRs live.