We tracked it down to the tick group TG_EndPhysics waiting for the ParticleSystemComponent to finish TG_DuringPhysics and this probably only happens with physics substepping enabled. At least that is what we suspect at the moment - should really be looked into by someone experienced with the thread management / task graph. As a workaround you can set the ParticleSystemComponent to use TG_PostPhysics - not sure how much performance this costs and if this can cause other problems, but it doesn’t stop responding anymore.
We didn’t experience any crashes, only freezes / non responding game after a random amount of time.