I would like to add that it would be very costly to re-validate everything. Imagine there is mass production. This “critical machine” might have produced thousands of products if not more.
I agree with most of you about the decision to re-evaluate depends on the investigation of what went wrong. Some machines break down after a certain amount of usage or its own “expiration date”. Since it’s an important one, there must be a backup. In that case, I would say if nothing’s majorly wrong (such as wrong specs, blueprint details…), re-validate a few products randomly (and according to protocol) before and after the machine replacement and go on with the production process.
If it’s a mass production line, the company must have taken this type of event into account as well, in that case, maybe following the protocol is best.