@jacobperron I was originally thinking to add this in a fork, but if the defaults aren't modified and flexibility is provided through environment variables, I guess we can merge the change on master.

BTW, this is ready for being reviewed. I opened it as a draft only because tests are missing.

@jacobperron I'm planning to write some tests before merging, to improve coverage.
We could kick a packaging build with this branch, I have done some manual testing and it's working fine.

Tests added, CI (testing rcl, test_rmw_implementation, rmw_connext_cpp, rmw_connext_shared_cpp):

• Linux
• Linux-aarch64 (failed because rti Connext isn't available in aarch64 in our CI)
• macOS
• Windows

Going in, thanks for the reviews @jacobperron !

@ivanpauno There is a chance that this introduced flakiness in the nightly Windows builds. Do you mind taking a look and seeing if it was this PR? These are the tests in question: https://ci.ros2.org/job/nightly_win_rel/1721/testReport/

@ivanpauno There is a chance that this introduced flakiness in the nightly Windows builds. Do you mind taking a look and seeing if it was this PR? These are the tests in question: https://ci.ros2.org/job/nightly_win_rel/1721/testReport/

I don't see how this PR can make that test fail, it's completely unrelated.
That test is failing to find that some nodes were set up within a timeout, and this PR cannot affect node discovery (I would be extremely surprised if it somehow does).

Most likely, the test was already flaky.

I didn't think so, but I also hadn't seen these tests pop up in recent history.

I'll see if I can reproduce, thanks!