Yeah this happens a lot, and in practice it's not a big deal. I noticed that when I redeploy, there's a nondeterministic flurry of errors in my system on the node that restarts. Finally I came to understand that this was race conditions. But it's honestly not a big deal. The system restarts anything in a funny state and moves on with its life. I feel like if you have something exquisitely race-condition sensitive that you aren't aware that it should be so, then you're not using OTP correctly.