the project lead, 5 October 2026: "why is it taking so long for pc2 and pc1s tasks to spin up?". The PCs have no inbound ports
and one miner is off the LAN, so one thread per app now long-polls the relay's public /wake with the last stamp
(GET <url>?since=<stamp>, held up to 45 s there). A changed stamp sends Event::Wake and the engine fetches the jobs
file at once (signature check unchanged); a wake during a fetch in flight fetches again right after it. The first
reply only seeds the stamp. Backoff 5, 15, then 60 s while the relay is unreachable, a 10 s floor between requests,
a full hold when the relay has no stamp yet: never a busy loop. One log line per wake, one when it falls back, one
on recovery. IGNEUM_APP_JOBS_WAKE_URL overrides the URL (set and empty: no waker).
CHECK_EVERY_S 600 -> 120: the poll is the fallback now; the first poll stays 40 to 60 s after start. An unchanged
file is no longer logged on every poll. Remote jobs off stops the waker too.
Unit tests: the stamp logic (seed, same, changed, empty), the backoff sequence and the recovery flag, the floor,
the URL and query. cargo test -p igneum-app: 60 + 22 pass; cargo build --release -p igneum-app: finished.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
c40aa303a6