With the current configuration, this shouldn't result in any major difference in scheduling behavior, as projects need to be 50s old to be eligible.

However, once we up the cohort count to 3 and reduce the minimum scheduling cutoff to 15s, we should shift in a few runs to 3 evenly distributed scheduling batches. If those prove to run quickly enough, we can (for US) increase to 4 or 6 cohorts if we think it'll help.

timedelta is safe here because the schedule itself doesn't determine what will run, only how frequently things can be triggered. The task itself has internal logic to look for last run timestamp and decide whether to run a cohort based on min and max allowable delay.