The root cause was a race condition in parallel Behave workers caused by
mutating a module-level singleton (_service) in
cleveragents.cli.commands.session. Concurrent cleanup in one worker could
reset _service to None while another worker was still using it, leading to
intermittent tell command test failures and exit code 1.
The fix patches the _get_session_service function directly in the affected
test steps (session_cli_coverage_boost_steps.py and
session_cli_uncovered_branches_steps.py) to avoid mutating the module-level
_service, and adds a reset call in features/environment.py's after_scenario
to ensure the singleton is cleared between scenarios, preventing stale
service instances from leaking across tests.
Closes#9121
HAL9000