disable caching to see the impact on runners
The performance impact (negative or positive) of the apt cache is negligible in the larger scheme of things. You can see this by comparing the timings of this job against other jobs that have ran with the cache. As an example, take those jobs:
https://gitlab.torproject.org/anarcat/tor/-/pipelines/16272/builds
and compare them with mike's:
https://gitlab.torproject.org/mikeperry/tor/-/pipelines/16246/builds
It's unclear why, but mike's caches include the build/
directory, which is ~640MB and means that the entire pipeline takes 5GB of disk space for the cache on the runner.
In comparison, the apt-cache/
directory is 80MB, which, in good conditions, would take less than a second to download on a gigabit link.
See also tpo/tpa/team#40493 (closed)