[BUG] Flaky integration test test_scenario_config_management on Rolling CI

[bburda]

Bug report

Steps to reproduce

1. Push any change to a branch with CI enabled
2. Wait for the build-and-test (rolling, ubuntu:noble) job
3. Observe test_scenario_config_management failing intermittently

Expected behavior

test_scenario_config_management should pass - it launches a single demo node (temp_sensor) and waits for the gateway to discover it.

Actual behavior

The test times out after 60 seconds with:

AssertionError: Discovery incomplete after 60.0s - found 2 apps, need 1.
Missing apps: {'temp_sensor'}, Missing areas: set()

Key observation: the gateway discovers 2 apps but none of them is temp_sensor. This suggests DDS cross-contamination from other tests leaking through ROS_DOMAIN_ID isolation on Rolling.

Root cause analysis

Three contributing factors:

1. DDS discovery is slower on Rolling - newer Fast-RTPS/CycloneDDS version with different timing characteristics. 60-second DISCOVERY_TIMEOUT is borderline.

2. ROS_DOMAIN_ID contamination - the test found 2 apps instead of the expected 1, but not the right one. This points to stale DDS participants from a previous test bleeding into this test's domain. The CMakeLists.txt already documents this risk:

   # Each test also gets a unique ROS_DOMAIN_ID to prevent DDS cross-contamination
   # between tests (e.g., stale DDS participants from previous tests leaking into
   # subsequent test's graph discovery).
   

3. CI runner contention - Docker-based Rolling CI shares resources, adding latency to DDS discovery.

Proposed fixes (ascending complexity)

1. Increase DISCOVERY_TIMEOUT from 60s to 90s (1-line change, lowest risk)

• File: src/ros2_medkit_integration_tests/ros2_medkit_test_utils/constants.py:39
• CMake scenario test timeout is already 300s, so 90s is well within bounds

2. Increase ROS_DOMAIN_ID stride for integration tests

• File: src/ros2_medkit_integration_tests/CMakeLists.txt
• Currently domain IDs are sequential (100, 101, 102...). Increasing stride to 2-5 would reduce DDS participant leakage between tests

3. Increase gateway refresh interval in tests from 1s to 2s

• File: src/ros2_medkit_integration_tests/ros2_medkit_test_utils/launch_helpers.py:98
• Reduces DDS middleware strain, gives more time per cycle

Environment

• ros2_medkit version: 0.3.0 (main branch, commit bdf6fe3)
• ROS 2 distro: Rolling (ubuntu:noble)
• OS: Ubuntu Noble (24.04) in Docker (GitHub Actions CI)

Additional information

• CI run: https://github.com/selfpatch/ros2_medkit/actions/runs/23308288985/job/67788427165
• Rolling is already marked continue-on-error: true in CI, so this doesn't block PRs
• Jazzy and Humble are not affected - their DDS versions have faster/more reliable discovery
• The same test passes consistently on Jazzy and Humble