5% bias calculated with the spiky load
$ python3 estimate_error.py
Starting Monte Carlo simulation with 1000 runs...
Each run simulates 30 days, divided into 8640 intervals of 5 minutes.
A events = 86136747, B events = 34253712
true = 0.92927688141684, agg = 0.9804034561428873, total_success = 1094375, total_fail = 1177663
Completed 100/1000 runs.
A events = 172610585, B events = 68464739