Assigning issues to cloud agents is quick, so pull requests arrive faster than reviewers can read them. Many need several rounds of feedback or a human commit to finish. Half-done pull requests pile up, review time goes to low-priority work, and the saved effort moves to the reviewer.
Starting a task costs one click, but reviewing one still costs a person's attention. Agents often need review iteration, and they repeat mistakes until the instruction file says otherwise.
Cap how many issues are assigned to agents at once, for example per reviewer. Assign only small, well-specified issues with a test command. Decide in advance when to finish a pull request by hand and when to close it, and close stale ones on a schedule. Put each repeated review comment into AGENTS.md or CLAUDE.md so the next pull request avoids it. Track merge and close rates per kind of task, and stop assigning kinds that rarely merge.
Toub warns that easy pull request creation can outpace reviewers. One person opened nine agent pull requests during a flight, and three were still open a week later. Most merged ones needed review iteration.
Boeckeler asks when a team would spend time finishing a partly working agent pull request for a low-priority task, and warns of a growing tail of unprioritised work.
The team says the most frustrating cost was the time spent salvaging failed attempts, and that it was hard to predict which tasks would succeed.