KV failure after removing entry point node for pre-MH server

Seeing an error from the SDK under this KV scenario, with pre-MH servers only: 

Operations: Continuous upserts and gets 

Change: Failover the entry point node.

Cluster:

1. kv <- removed, (used as the connection string)

2. kv,index,fts,n1ql 

3. kv,index,fts,n1ql

4. kv,index,fts,n1ql

Error: Seeing a couple errors but it settles on a repeated authentication error 

2020-04-01 04:07:01.7551||Couchbase.Core.IO.Authentication.ScramShaMechanism|ERROR|Authentication failed. Cannot authenticate the user. Reason: AuthenticationError

Graph results: http://sdk-testresults.couchbase.com.s3.amazonaws.com/SDK-SDK/CB-6.0.3-2895/FoEptEject-HYBRID/03-31-20/079249/302532bddf98464d9978469df49a52ca-MC.html

Full logs attached -> 

 , and for the run logs http://sdkbuilds.sc.couchbase.com/view/.NET/job/sdk-net-situational-release/job/SDK3.0-Release/job/net-centos-alice/3/consoleFull (which has some more of the sdkd logs in context, specifically if you look after "Master node being removed from cluster. Switching", you'll see where the errors begin).

Note: This issue causes 6 test failures, all of which involve removing the EPT node whilst running KV ops (upsert, get and subdoc). This is the most basic of the test scenarios where the EPT node is just ejected.