There are situations where a bucket connection pool gets closed and refreshed.
In this case getting a connection from a pool will return a "no connection pool" error until connections are moved from the old pool to the new pool, which will probably take a minute.
There is little point in doBulkGet() in doing a retry because it will likely go through all the retries very quickly and fail, and hog the CPU in the process.
noErrPool should abort rather than retry, or at the very least back off.
- relates to
MB-39804 Query service returns Error performing bulk get operation