Many years ago I have been wondering if long builds could be accelerated by just observing how the processes in the build behaved and caching and replaying the outputs of the well-behaving ones.

Apparently this can work and perform quite well! (... after a ton of work :-)) The tool we built is available on the link above for local testing and to be used as a GitHub Action.

Please give it a try and let us know what you think.