As mentioned in this comment from PR #591 and PR #564, failing to end a session in a test where an exception is thrown could cause future test methods to encounter delays in execution. This is mainly an issue with transaction tests, where an exception's backtrace can hold a reference to the session object and prevent it from being garbage-collected after the test function exits.

If the test class is able to track sessions that are used within tests (e.g. creating them through a convenience method instead of directly calling Manager::startSession() within the test), we should be able to hold a reference to them and ensure they are explicitly ended in tearDown(). This would allow us to remove try / finally code blocks in various tests that currently work around this issue.