3 ms·

I agree, as long as you store everything permanent on a centralized database then being able to fix a node in a cluster is just a card swap.