[apostasie]

A lot of work has improved stability since I last tried this.
Though we still have some flaky tests in the "non-retry" bucket, we need more info at this point on the state of the "retry" bucket that is likely hiding more issues.

This is a debugging PR at this point, not meant to be merged.

[apostasie]

Ok... this is not a pleasing sight...

[apostasie]

First run surfaced:

• [LEGACY] [CI]: TestRunBindMountBind is not working well #4144
• [CI]: OCI runtime start failed: cannot start a container that has stopped: unknown #4143
• [CI]: buildkit Internal: not found #4142
• [CI]: Syslog is flaky: TestSyslogNetwork TestSyslogFacilities #4141

and unfortunately, yet again IPFS digest not found.

[apostasie]

Second run brings:

• one count of [CI]: OCI runtime start failed: cannot start a container that has stopped: unknown #4143
• one count of [CI]: TestComposeUpWithExternalNetwork network failure #4146
• several counts of [CI]: buildkit Internal: not found #4142
• one count of [CI]: TestLogsWithoutNewlineOrEOF flaky #4147
• several counts of compose ipfs failure to communicate with database
• multiple counts of TestIPFSCompNoBuild content digest failure

[apostasie]

Third run brings:

• one count of [CI]: OCI runtime start failed: cannot start a container that has stopped: unknown #4143 on TestComposeUpWithBypass4netns
• four counts of [CI]: buildkit Internal: not found #4142 on the same run <- we have a serious problem here

[apostasie]

Pausing this for now.
There is no point in poking this more - we have enough issues to fix from the first few runs...