Monitoring baselines and safe restarts

In the first stable week, record with timestamps: accepted hashrate (pool side), per-board temps at a known intake temperature, fan speeds, wall watts, reject rate. That table is your diagnostic reference forever — 'temps are high' is a feeling; 'board 2 is 9°C above its March baseline at the same intake' is a fact.

• Pool-side hashrate (truth) vs miner-side (optimism) — a growing gap means rejects/stales.
• Alert thresholds: hashrate −20% for 30 min, any temp over manufacturer limit, pool unreachable 15 min, reject rate >3%.
• Log where future-you will find it: a spreadsheet is fine; discipline beats dashboards.

• 1. Graceful: the miner UI/API 'stop mining' → wait for fans to idle → start. Preserves logs and the filesystem.
• 2. Soft reboot via UI (like a computer restart).
• 3. Power cycle at the socket — last resort; repeated hard cuts are not a control strategy, they are wear.
• 4. After ANY restart: confirm shares flowing to the right worker before walking away.

Pool cards on project pages expose worker-level stats endpoints — the same feeds worth piping into your own alerting.