Add a missing test case to assert that a driver does not retry a read when in a transaction. This would have prevented
The requirement is in the retryable reads specification in the when not to allow retry section.
Who is the affected end user?
How likely is it that this problem or use case will occur?
This seems like an easy bug for a driver to have or introduce, and it's hard to detect since it depends on a server failure at just the right time
If the problem does occur, what are the consequences and how severe are they?
Applications won't get the benefit of the convent transaction API's automatic retry mechanism if a driver has this bug
Is this issue urgent?
Yes, since we want to fix the Java driver bug and need a test to assert correct behavior.
Is this ticket required by a downstream team?
Is this ticket only for tests?
It's a missing test of required behavior