By running prefork, you still risk two things.
The first is that if you don’t disable Python interpreter initialisation in the Apache child worker processes, that will slow spin up of prefork worker processes, plus still cause a lot more memory to be used that necessary.
The second is that how Apache dynamically adjusts the number of child worker processes causes a lot more worker process churn with prefork. This is made much worse when Python interpreter initialisation isn’t disabled.
Using worker MPM is still preferred because in most cases people still don’t adjust the MPM settings appropriately and so worker MPM is more forgiving to things not being done properly. You don’t even describe what MPM settings you are using, yet they are still quite important.
For further research see/read:
Graham