Q: What does `git gc` do, and when should you run it manually?

git gc (garbage collection) performs housekeeping on the Git object database:

• Packs loose objects into packfiles — individual objects in .git/objects/ are compressed into efficient .pack files with delta compression.
• Removes unreachable objects — commits, blobs, and trees that no ref (branch, tag, reflog) points to. These are typically from amended commits, rebases, or deleted branches.
• Packs refs — consolidates individual ref files in .git/refs/ into a single packed-refs file.
• Prunes old reflog entries — entries older than 90 days (reachable) or 30 days (unreachable) by default.

When Git runs it automatically: Git triggers gc --auto after certain operations (e.g., when there are more than 6700 loose objects or more than 50 packfiles). You rarely need to run it manually. When to run manually: Aggressive GC: This spends more CPU time on delta compression for a smaller packfile. Only useful occasionally (e.g., after importing from another VCS). Don't run it routinely — the default compression is good enough and --aggressive is slow. Caution: --prune=now immediately deletes unreachable objects. If you're recovering lost commits via reflog, run recovery *before* GC.

git gc --aggressive --prune=now

• After large history rewrites (filter-repo, mass deletions) to reclaim disk space.
• When the repo feels slow and .git/ is unusually large.
• Before creating a bundle or archive — ensures maximum compression.