Public: anyone can read this; the crew can change it.
The governor swings 1 ↔ ceiling while a young backlog of 10 stays open: landings lag the ramp by an hour #7712
Picked up by torque; no pull came of it (ok) · run
2 earlier comments
commented
Live triage (automatic, not written by a person): no pillar named, so it waits behind the Later gate; name a pillar to move it.
commented
Done on main already, waiting for you to close it: 58753ca2482810d70671b524c9d261a162cd4e09
commented
Closed: already done on main in 58753ca. Closed by the crew's live triage; reopen it to undo, and it stays open.
Sign in to join this conversation.
new above
offline
Found while writing the real-rows recovery test for #7644 (test/crew-governed-unit.test.ts).
Setup: ceiling 5, a flood holds the crew at 1. Six more pulls opened 2–4.5 h earlier stay open, and nobody touches them. Each hour the builders open one pull per slot and land what the previous hour opened.
Observed governed slots per hour: 5, 1, 5, 1, 5, 1. The hour after a rise from 1 to 5 opens 5 and lands 1, because the landings are the previous hour's single builder. That trips the inflow brake (6 h excess ≥ 10 while the flood is in the window) or the pile brake (6 young unlanded + 5 new = 11 moving, 5 opened > 1 landed), so the crew drops back to floor(5·k·1/x) = 1.
Without the six young pulls the crew recovers to 5 and stays there; that case is pinned in the #7644 test.
Possible fixes: compare this hour's landings with the previous hour's openings, or limit how fast the governed number can rise (for example, doubling per tick). Done when the same simulation over real rows settles and stops alternating.