How exactly is it determining the "most relevant" commits? Is this just filtering out commits that don't touch the current file(s)? Or some kind of scoring metric?
1. The branch isn't following a remote branch that's been deleted (which often happens after merging or deleting a pull request).
2. The branch has a common commit with the primary remote branch (usually origin/main or origin/master) from the last 2 weeks. You can see older commits & branches by clicking “load more commits.”
Then, it displays all the commits from the top of that branch down to where it joins with the main branch.
If the HEAD commit meets the primary branch at a commit more than 2 weeks old, then the 2 week range is extended (for all branches) as well.