As part of PERF-5529 we investigated some changes to flow control that could be worth improving in the future:

• Investigate changing flow control to use the lag between the primary's lastApplied and last majority applied instead of primary's lastApplied and lastCommitted (after SPM-3489 lastCommitted means the last majority written) This may not be necessary anymore after SERVER-92065. 
• Investigate tuning default parameters like flowControlMinTicketsPerSecond and flowControlSamplePeriod. We found that the default flowControlMinTicketsPerSecond of 100 meant that with batch inserts of many operations, flow control didn't end up decreasing secondary lag at all. We also found that with the default flowControlSamplePeriod of 1000, this meant that we didn't sample often enough, which meant that flow control was getting disengaged even when there was lag still occurring. I'm not sure how the defaults were chosen initially but it's possible with SPM-3489 that they should be reevaluated.
• Investigating having flow control take into account the number of operations in a batch
• Increase testing for flow control - Currently there is a C++ unit test and one integration test but it would be nice to add more integration tests if possible. It also seems like we have some perf testing with flow control enabled from PERF-1818 and PERF-1841 but I'm not sure if any of these workloads have enough secondary lag to result in flow control actually getting engaged in slowing down the primary so it might be nice to have more perf testing with more secondary lag to test flow control