nit: Just to stay on the safer side, move it closer to the command that actually does the build. As it is, everything will work ok, just to fool-proof ourselves from cases where -e is accidentally removed.

I don't think this makes sense in the Python script, because it has a bunch of different builds in conditional blocks, so the only way to consistently get all stats is to put it at the end. I made the change in the protoc build script. There aren't any meaningful stats in the other scripts, so I removed the command there.

We may also want to compress it to something like

[ -x "$(command -v ccache)" ] && ccache --show-stats || true

I tried this, but shellcheck rejected it so I went back to the form I started with.

The pear package command typically only bundles source files into an archive and does not perform any C/C++ compilation. Since no compilation occurs within this script, reporting ccache statistics here is likely unnecessary and will not provide useful signal regarding build performance.

To reduce boilerplate across multiple scripts, consider using a more concise one-liner. Additionally, note that this script does not source tools/internal_ci/helper_scripts/prepare_ccache_symlinks_rc. Unless ccache is already configured to wrap the compiler in the environment's PATH, these statistics may not reflect any actual build activity.

Consider using a more concise one-liner for reporting optional statistics to keep the script clean.

Consider using a more concise one-liner. Also, similar to the protoc artifact build, this script lacks the explicit ccache initialization found in build_artifact_python.sh. If the compiler is not correctly wrapped via PATH symlinks, these statistics may remain empty.

This script only performs file copying (cp) of pre-built artifacts and does not involve any compilation. Reporting ccache statistics here is unnecessary as it won't reflect any activity from this specific build phase.