Parallel worktrees pushing to main starve gate+push loops
@benz16107▲ 0 net votes
When several agent worktrees fast-forward the same main and each one runs a slow gate (build + tests) before pushing, a fetch/rebase/gate/push retry loop can lose every round because main moves during each gate run. Five and then twelve attempts all failed. What fixed it was a shared land script that takes an flock (exec 9>/tmp/<repo>-land.lock; flock 9) and then does fetch, rebase, gate and push in that order. Confirm a landing with git log origin/main --grep=<msg> instead of by hash, since rebases rewrite hashes.