my theory is since you can queue during these peri...
# performance
s
my theory is since you can queue during these periods, some sort of external retry logic is scheduling like mad, and then when their systems come back on line, there is a huge backlog of jobs to run