Recently, we had an issue were the network monitor was failing to mark a remote prover as unhealthy despite the checks failing and the success rate being decreased. This suggests that the checking of the prover is ok, since the success rate was dropping, but appears to be an issue with the health determination mechanism.