Verification run for f86b604: four pods, full page heap throughout, the
fixed binary confirmed by hash in the first line of the log. Twenty valid
samples - every one of them four pods on the podium and a full teardown -
and no fault.
Before the fix the rig trapped twice in eleven instrumented launches, about
one in five. Twenty clean at that rate is roughly a one-in-fifty
coincidence, which is as much as twenty launches can say and enough to call
it held.
What this does NOT settle is the crash the rig was built for. The original
playtest dumps are a double-delete in ~JointedMover under
Application::Shutdown, and that stack has not appeared once in thirty-odd
instrumented launches. The bug that was caught and fixed is a real
use-after-free on the same path - a Player following a vehicle that had been
fried - and it may well be the same root, since a VTV freed while others
still hold it would be deleted again at Shutdown. But nothing here
demonstrates that, and the rig stays.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
CydandClaude Opus 5
b3df9f0804