Do we know what introduced this issue? Can we add that details to the PR description to help future readers?

Do we know what introduced this issue? Can we add that details to the PR description to help future readers?

Noticed in-proc backport not required. So, the DI Changes triggered this?

Do we know what introduced this issue? Can we add that details to the PR description to help future readers?

The pattern previously in place was fragile, and created an opportunity for services to be consumed after disposal. The flow of events was:

1. API was called
2. Restart suppressed
3. Files written (normally, this would trigger a JobHost restart, the suppression prevented that from happening)
4. Restart resumed, processing pending changes
5. API call returned

When step 4 takes place, the JobHost will be terminated, and the JobHost scoped container will be disposed, causing all associated services to be disposed as well. In in-proc, although there is some evidence of the problem happening, so a backport should probably be considered, the restart didn't occur fast enough (or there were no additional service requests) to impact the response handling, with the changes in the default host, the JobHost disposal is taking place before the response is handled and no longer needs to consume services.

This was a race that was in place from the time that cod was introduced, just not something that was exposed with the previous host.

Hope the above answers the second question as well.

Did you miss the release notes file?