Comment 1 for bug 1749081

The worker is started with a millisecond timeout, with the assumption that it is up before the connection attempt. Sometimes it is not.

Simple to reproduce consistently - just wind out the timeout and it fails every time with the message observed in the CI failure.