The start up cache gets filled at most ~60s after start up and improves subsequent cold view start ups by at most 508ms on a Moto G5 (source). Most users will experience start ups with this cache so we should be measuring performance with this cache: let's update both of measure_start_up and backfill.py to take it into account.
Note: a bug BZ-1728461 prevented the start up cache from taking affect until > 6 hours after first run so we haven't often seen this discrepancy locally.
The start up cache gets filled at most ~60s after start up and improves subsequent cold view start ups by at most 508ms on a Moto G5 (source). Most users will experience start ups with this cache so we should be measuring performance with this cache: let's update both of measure_start_up and backfill.py to take it into account.
We might just be able to change this line.
Note: a bug BZ-1728461 prevented the start up cache from taking affect until > 6 hours after first run so we haven't often seen this discrepancy locally.