Better compression algorithm for artifacts and caches
Today both artifacts and caches in CI use ZIP with a very bad implementation (uses single core).
ZIP is inappropriate archiving+compression method for text artifacts because it compresses each entry in archive separately, which makes archives sometimes 100% bigger than compared gzipped TAR or ISO files. On the other hand, a lot, if not most binary artifacts already use some form of compression, so ZIP will not improve (or, likely, even inflate) the size of artifacts.
On top of it, the ZIP implementation chosen is very slow, because it uses only one CPU core.
On top of it, it is possible to mount an archive, both ZIP and TAR, or ISO, with, say,
archivemount, but for some reason you are not doing it (you are extracting the archive before doing anything else).
All this said, I believe it would be great, if it was possible to choose from at least these options for archiving the cache and artifacts:
none. Nice to have:
tar+bz etc... Also, from cursory reading, 7z implements a speedier ZIP version (with threading support), so, perhaps, replace the one you are using now with this one.