Tests with MultiThreadedExecutor time out on some CI instances

[jacobperron]

Bug report

Required Info:

• Operating System:
  • Any
• Installation type:
  • Source
• Version or commit hash:
  • master
• DDS implementation:
  • N/A
• Client library (if applicable):
  • rclpy

Steps to reproduce issue

Unable to reproduce locally on Linux.

Example CI job failing: https://ci.ros2.org/job/ci_linux/6113/

Disabling this test resolves the issue:

rclpy/rclpy/test/action/test_client.py

Lines 257 to 279 in b6c6ca8

	def test_send_goal_multiple(self):
	ac = ActionClient(
	self.node,
	Fibonacci,
	'fibonacci',
	callback_group=ReentrantCallbackGroup())
	executor = MultiThreadedExecutor(context=self.context)
	try:
	self.assertTrue(ac.wait_for_server(timeout_sec=1.0))
	future_0 = ac.send_goal_async(Fibonacci.Goal())
	future_1 = ac.send_goal_async(Fibonacci.Goal())
	future_2 = ac.send_goal_async(Fibonacci.Goal())
	rclpy.spin_until_future_complete(self.node, future_0, executor)
	rclpy.spin_until_future_complete(self.node, future_1, executor)
	rclpy.spin_until_future_complete(self.node, future_2, executor)
	self.assertTrue(future_0.done())
	self.assertTrue(future_1.done())
	self.assertTrue(future_2.done())
	self.assertTrue(future_0.result().accepted)
	self.assertTrue(future_1.result().accepted)
	self.assertTrue(future_2.result().accepted)
	finally:
	ac.destroy()