I think no one has ever encountered this type of inconsistency. Though i am still researching for why this issue occurs randomly, steps i have performed till now are as below

1. Initial assumption was that some reports in production might be causing this inconsistency and leading to service restart. We did check the event and audit logs and identified the reports which were being accessed near to service restart time and what actions were performed by end user. These actions on sample reports (which are heavy reports) were performed again in production by the developers to pinpoint the cause of failure, however there was nothing unusual being observed. However we spent a lot of time analyzing different reports but this didn't give us concrete answer to the error being logged on the server and affecting downtime. We did maintain the reports which were audited

2. On analyzing the logs we could see some lines being logged with regards to leavers and these were marked as warning messages. These were either server objects or bookmarks (primarily) which were being logged. We could not relate these to service restart, but as a check we did check all the reports in QMC under server documents and checked what objects / bookmarks were available against leavers and cleared all of them. Not sure which bookmark was causing the issue but post the cleanup we have not seen a single service restart with this error or warning being logged on production server.

Just posting it here to understand if the community has observed any of the above and what could be the reason for leavers bookmark being applied during report refresh causing this warning and error logs.