Closed WrathfulSpatula closed 2 years ago
The last commit might have fixed the issue, judging by CI and local test runs, but the code is in no worse state, and the refactor here is better, rather. Merging, and I'll continue to apply scrutiny if we haven't nailed it.
I locally recreated the CI failure from today. For some reason about the shard cache, in the process of separation, with
AI()
/IAI()
1qb state preparation checks involved, this is at least a reliably passing workaround that also makes sense to me, instead.