Any downtime during a workflow run on hosted windows runners possible?

[sschabbach-mse]

We observed a high variance in performnce running our builds on github hosted windows 2025 runners. Especially one of our tests is expecting an change event from a mongo database:

Passed ChangedTest [8 s]
Passed ChangedTest [7 s]
Passed ChangedTest [7 s]
Passed ChangedTest [831 ms]
Failed ChangedTest [3 m 1 s]

As you can see, the event is received in less that one second (1 test run), mostly it takes about 10 seconds and in the last test run the event is not received after 3 Minutes!!!

We can't image that the performance can be a reason and would like to ask if it is possible that we can have downtimes for any reason within a single job execution (such as moving to another infrastructure), that can explain these behaviour.

Kind regards