The [user32:sysparams patterns page](https://test.winehq.org/data/patterns.html#user32:sysparams) shows plenty of failures with Windows 10.
Running [a single machine](https://testbot.winehq.org/JobDetails.pl?Key=151682) test did not show such a failure. Therefore the idea is there is some race condition, which shows up when testbot has more load, like in the batch runs.
While there was still a batch run active, [I received with a null change](https://testbot.winehq.org/JobDetails.pl?Key=151697) the failures quite reliable.
[Adding some Sleeps](https://testbot.winehq.org/JobDetails.pl?Key=151698) seems to improve the situation.
This Sleeps would add 4 seconds to each test run.