User login is hanging with mongo auth credentials

As part of the work, we are bringing a three member replica set (Primary, secondary & Arbiter) and enabled the mongoAuth in that. As part of the Auth, created two users (User1 & User2). User1 has mapped with more roles as “userAdminAnyDatabase,readWriteAnyDatabase,clusterAdmin,dbAdminAnyDatabase,restore,backup & hostManager”. Where as User2 has mapped with “readAnyDatabase,clusterMonitor and backup”. After creating both the users, we are able to login to the primary & seconday replica’s with the both users with out any issues.

On top of this we are running 30Hrs traffic test, where replica set’s are stressed by updating the information. After the 30Hrs we have observed, User2 login is getting stuck. The command used for this is,

mongo -u User2 -p password --authenticationDatabase admin <hostname>:<portnum>

Since the above command got hanged, logged into the secondary with the another user (User1) and from that observed that, it is getting hanged even for the command “show users” as well. When we verified the current running operations by using “db.currentOp()“, Saw multiple operations running for a longer time like below(check the secs_running). Later I tried kill one of the operation by using “db.killOp(opid)” but that also looks in pending state, which is shown in ““killPending” : true”

{
   "host":"sessionmgr12:27719",
   "desc":"conn92342",
   "connectionId":92342,
   "client":"192.35.11.82:56230",
   "active":true,
   "currentOpTime":"2020-07-22T11:06:49.792+0000",
   "opid":978477321,
   "killPending":true,
   "secs_running":NumberLong(17632),
   "microsecs_running":"NumberLong("   "17632917128"   ")",
   "op":"command",
   "ns":"admin.$cmd",
   "command":{
      "saslStart":1,
      "mechanism":"SCRAM-SHA-1",
      "payload":BinData(0,
      "biwsbj1yZWFkb25seSxyPXdRVlNhdjdP"      ")",
      "$db":"admin"
   },
   "numYields":0,
   "locks":{

   },
   "waitingForLock":false,
   "lockStats":{

   }
}

We are using mongo version 3.6.17 on top of the Centos8. As part of this process, have we missed any thing here ? Or else do we need to tune anything on the mongo users ? or else is it a known issue in the 3.6.17 version ?