Most replays are never noticed. Most bugs don't require immediate replays by all nodes for the system to remain functional. The frozen chain issue we had recently and this issue with the hard fork were extreme examples. Mostly replays are done on backups and then brought in seamlessly (which is why most people don't notice them).

I do hear you, though. Speeding up replays will help minimize downtime in situations like this. It's certainly being talked about and improvements in both the hardware possibilities available and in the software may help make this a reality.