Sometimes trajectory_follower's test fails with ARM64

[kenji-miyake]

Checklist

• I've read the contribution guidelines.
• I've searched other issues and no duplicate issues were found.
• I'm convinced that this is not my fault but a bug.

Description

See https://github.com/autowarefoundation/autoware.universe/runs/5860423885?check_suite_focus=true.

Expected behavior

CI passes.

Actual behavior

CI failed.

Steps to reproduce

Run the build-and-test-self-hosted workflow several times.

Versions

ARM64

Possible causes

No response

Additional context

Previously, it passed if I re-run the workflow. So I guess it's just flaky.

[kenji-miyake]

@TakaHoribe @taikitanaka3 Do you know any causes?

[maxime-clem]

Previously flaky tests were sometimes caused by interference between tests that use the same topics running in parallel.
I have opened a draft PR #661 to switch the tests of the trajectory_follower_nodes to run in isolation but it is hard to verify if this solves the issue.

[kenji-miyake]

@esteve cc @mitsudome-r @xmfcx Are you familiar with this issue? I remember you've worked on this kind of task in Autoware.Auto.
If so, could you give me your thoughts? 🙏