After every release a cell child cannot wake for ten minutes: the boot opens every hosted child and fills FORGE_CHILD_AWAKE_MAX ("the forge is full") #10502

Seen 2026-11-10 02:00–02:08Z. scottishbob's cell had been stopped and was answered 429 "the forge is full, try again shortly" for 8 minutes, starting 3 minutes after 0.7.506 booted.

Cause. bootHostedChildren (src/spawn/host.ts) opens every finished hosted child at boot so that its apps come back. There are 8 hosted children today (grokbot-dev, grokdemo, crewshop, gamer, mark, homie, real-estate-managemtn, grokbot). awakeCount counts each of them, plus the cells, against awakeCeiling (FORGE_CHILD_AWAKE_MAX, default 8). The count therefore sits at or above the ceiling until the idle sweep sleeps them 10 minutes later. Releases run about every 30 minutes, so for roughly a third of the time no sleeping child can wake. In the meantime its crew is frozen and its owner sees "the forge is full".

Wanted (pick the simplest that holds).

• A child that boot opened and that nobody has asked for does not count against the ceiling, or boot sleeps it as soon as its apps are back. Or:
• The boot opens only the children that were awake when the previous process left, recorded in a small spawn/awake.json on drain.
• Either way, a person's wake is never refused because of forges nobody is using.

Done when. A test boots a parent with 8 idle hosted children and one sleeping cell child. A request to the cell child wakes it at once, and is not answered 429.

Part of P4 #590.

Parked by the crew: the check stayed red through 2 automatic tries (pull #10504 closed; its branch torque/10502-after-every-release-a-cell-child-cannot at 325f1a75f26a keeps the work so far). Planner: split this smaller, or rewrite what done looks like; a person can also say what to try.