I'm able to reproduce this bug by introducing a delay after the check of protocol version but before the schedule of updateTerm several lines later, so that term will be updated after the pv=0 reconfig in executor. As a result, the primary steps down because of the PV1 term in getMore request, even though it's in PV0. The time window of this bug is only several lines of code, but it did happen on Evergreen.