For context, as runaway tasks were executed, this caused the infrastructure container CPU utilization to spike (green line, yellow line is the ephemeral service container) due to all the database operations. This caused multiple issues including failure of multiple tasks and even some service calls (mostly due to database errors related to blocking calls), caused task execution times to be excessive (new day tasks were running in the realm of 30 minutes, whereby they usually take about 30 seconds) and eventually crash entirely.
The root cause was a race condition in the way that the new day task handled arcade world (since it advances by 7 days rather than 1), causing more new day tasks to be scheduled than were supposed to. This was not the issue in previous event manager due to a check for a scheduled event that prevented it, but this check did not work as expected in new flow.