Triggers vs schedules: why polling every ten minutes backfires
The most common thing I see in a new Superagent by
@Base44 setup is a schedule that runs every ten minutes. Check my email every ten minutes, check the repo every ten minutes. It runs, which is not the same as working. It runs six times an hour to find out that nothing happened, and every one of those runs spends message credits.
The cost is the smaller problem. If the connected service starts rate limiting you, runs stop finishing and start queueing behind each other, so the thing you wanted at 9:04 arrives at 9:40 with the backlog behind it.
Use a schedule when the work actually belongs to a clock. A morning brief, or a report that goes out every Monday.
Anything that should happen because something happened belongs on a trigger. Triggers fire when your data changes or when there is activity in a connected tool, so the agent moves at the moment of the change instead of ten minutes after it. Same outcome, a fraction of the runs, and nothing piled up behind itself.