GetContainerItemInfo returns nil for some occupied slots during server
lag, causing Bagnon to draw the slot empty. Earlier retry-polling
approach caused massive lag because every legitimately emptied slot
also matched the "previously had item, now nil" condition and ran ~30
live reads over 3s per move.
New approach: on a nil-after-occupied read, if we're within
GRACE_WINDOW (0.5s) of the last good draw, keep the prior draw and
queue ONE re-check at the deadline — no polling. A genuinely emptied
slot resolves at the deadline; a recovered slot redraws with correct
data. lastGoodTime is set only on non-nil reads, so persistent lag
eventually accepts the empty.
florian.berthold
c955937b12