Live and learn is what I think the take away of this story is all about... I had a server fail dec 25 mid morning. It caused failures in away I had thought about before because instead appearing completely dead it was alive enough to not let go of any tcp connections. For the critical component in question, I didn’t have the correct timeouts in place... so as the single operator I was fortunate that my wife was also my co founder and so was a bit more understanding.