Any reason not to do that? The FVWM race condition is making the tests fail quite often now, because dinput device acquisition is faster than it used to be. FWIW I opened https://github.com/fvwmorg/fvwm/pull/101 to fix it upstream, but it doesn't look very active.