While upgrading mongodb from mongo 3.2 - mongo 3.4 on amazon linux machine mongo upgrade process on secondary servers is taking 20mins on each node, while on primary and read replica node  mongo upgrade process completed in 2min. 

The time took by upgrade process on secondary servers is 10mins while running transaction test and other 10min after transaction test was successful.

What can be the possible reason for this time?  The upgrade process was done on non-prod environment having half n hour of oplog window. On production environment on what basis can we predict the possible time for this process?? there can be the chances node going into recovery state on production environment if process takes an hour or so?