Is it all existing pngs, or a subset?

All, however, it would just save off a byte or two for a few.

I really don't think we want to exclude all existing images, that will severely reduce the usefulness of the new check. After all, the most likely changes are going to be to existing images to update them.

Does anyone know why we use --all-files in the CI pre-commit check? Surely it would be better to make use of --from-ref and --to-ref?

Alternately, we should just take the hit once to enable the check and optimize the snot out of everything in this PR with no exclude list.

Optimising exisiting PNGs has some benefit, but results in a larger repo size. The only user-visible effects should be in the documentation, where we'd preferably use SVGs to PNGs (especially for diagrams etc).

Personally, I think this check would probably cost more than it saves (at least on a macro scale), but I don't have a strong opinion, I don't use pre-commit.