connector_detatch Test Occasionally Fails on Windows

[CoolSpy3]

Describe the Bug
I've noticed on a few PRs that the connector_detatch test sometimes fails when run on Windows with the message Connectors should not be detached when applying a force < 2.1 N. This usually goes away if the test is rerun. Locally, I've been able to reproduce this on the master branch, and while I haven't been counting, I'd estimate that the failure rate is ~5% (give or take). When it fails, it appears to do so at 0.096 s, which corresponds to the first time the relevant check is run (after 3 timesteps). This probably means that the connector fails to attach properly at the beginning.

It should be noted that in my (fairly limited) testing, I was unable to reproduce this failure in Webots directly (either normally, in fast mode, or without rendering); it only occurred when running the suite (python tests/test_suite.py tests/physics/worlds/connector_detach.wbt --nomake). Thus, it could be related running Webots in minimized mode. Alternatively, it could just be bad luck due to the fairly low incidence rate.

Expected behavior
The test should always pass.

System

• Operating System: Windows 10
• Graphics Card: NVIDIA GeForce GTX 1050 Ti

[ShuffleWire]

For what it's worth : In my experience, the connector is sometime non repeatable : after a reset, it sometime fails to start as locked (or unlocked, depending of the use case). It should be something like 5% of the time as well. Have a look to #5889
However, I think that in the pipeline, the simulation is fully reloaded, and in that case I didn't get any issue, so idk...