Summary

When doctor reports a stale gbrain-cycle lock, it advises gbrain sync --break-lock — but that command can only address gbrain-sync:* locks. There is no shipped CLI handler for gbrain-cycle locks, so the only way to clear them is direct SQL into gbrain_cycle_locks.

Repro

1. Have a dream cycle die mid-run (process crash, OOM, sigkill) so the gbrain-cycle row stays in gbrain_cycle_locks past its TTL
2. Run gbrain doctor. The stale_locks check correctly detects it and prints:
   gbrain-cycle (pid <N> on <host>, age <age>) → gbrain sync --break-lock
3. Run the suggested gbrain sync --break-lock
4. Output: Lock gbrain-sync:default is not held (nothing to break).
5. The cycle lock is unchanged, and subsequent cycle_freshness checks continue to FAIL

Root cause

src/commands/doctor.ts:1670 hard-codes the hint:

const breakHint = source ? `gbrain sync --break-lock --source ${source}` : `gbrain sync --break-lock`;

src/commands/sync.ts:556 scopes --break-lock to sync locks only:

`gbrain sync --break-lock --source ${lockKey.slice('gbrain-sync:'.length)}`

There is no analogous handler for gbrain-cycle (or any other lock prefix).

Tested workarounds that did not work

• gbrain sync --break-lock — wrong scope (handles gbrain-sync:* only)
• gbrain sync --break-lock --source default — same
• gbrain sync --break-lock --source all — same
• gbrain dream --break-lock — flag silently dropped; just runs a cycle (same shape as gbrain dream silently drops --source flag, breaking cycle_freshness loop doctor recommends #1452)

Suggested fixes

1. Doctor: branch the hint based on the lock_name prefix (e.g. gbrain-cycle → suggest the cycle breaker)
2. CLI: ship a handler. Either:
   • gbrain dream --break-lock (route by command that owns the lock kind), or
   • gbrain admin break-lock <lock-name> (kind-agnostic)
3. Both should reuse the same liveness + force-break semantics as sync.tss D3 implementation (PID check, TTL window, cross-host safety, --force-break-lock opt-out)

Environment

• gbrain v0.41.14.0
• Hit on gbrain-cycle after a dream cycle died at 09:00 UTC; lock TTL expired at 09:30 UTC; PID 3396714 was dead per ps -p at 21:30 UTC
• Stale lock blocked all subsequent dream/cycle runs (cycle_freshness FAIL: "last cycled 30h ago")

Related

• gbrain dream silently drops --source flag, breaking cycle_freshness loop doctor recommends #1452 (dream silently drops --source) — same pattern of flags being dropped
• doctor advertises agent.use_gateway_loop=true as fix, but config CLI rejects it #1454 (doctor advertises config option that does not exist) — same pattern of doctor hints not matching shipped CLI surface