KV throughput against Trinity drops during rebalances

The KV performance against 7.6 when the cluster undergoes rebalances seems to be quite poor compared to previous server versions and often comes with a number of timed out operations.

Here we're using FIT-SIT against Capella 7.6, doing a 3 -> 5 node scale:

https://performance-sdk.couchbase.com:8080/situationalSingle?situationalRunId=212f85d4-625b-4f89-b435-9642cd1e252e&runId=b61880ea-b696-4f44-a447-199f05b57a52

You can see a period of 1.5-2mins where the throughput has dropped a lot - basically to 0. And there are 72 KV errors in this time (the performer needs an improvement to record the actual time these error happen, hence them not showing on the graph).

Compared to Capella 7.2 with the same test:

https://performance-sdk.couchbase.com:8080/situationalSingle?situationalRunId=9495c826-01d5-4af3-b4cf-88dbb716f4ab&runId=6d3fb2e0-26a1-46ce-8557-f06bc068e540

No errors, and very little throughput disturbance. 

This seems to happen for any scale up or down, and we see similar things for 7.6 on SDKD, though the performance does generally seem to be better with sdkd.

Note that in all test cases the SDK does eventually re-establish the same throughput as before the cluster change.