Bug report

• Operating System: Ubuntu 20.04
• ROS2 Version: foxy
• Version or commit hash: 2e18dd1
• DDS implementation: Fast-RTPS

Steps to reproduce issue

• Use any Behavior Tree Action Node ( BT::ActionNodeBase )
• Call an action with lower-than-needed timeout.

Expected behavior

• The action node itself should return a BT::NodeStatus::FAILURE whenever a problem occurs.
• Whenever there is an error, clear and detailed error logs should be provided.
• The ros2 node holding the BT engine ( like bt_navigator ) should not die
• The BT engine should keep running the tree (for instance, running a recovery node )

Actual behavior

• The whole ros2 node holding the BT engine crashes ( bt_navigator ), thus no recovery node is execute but rather the whole BT execution is suddenly halted
• No meaningful logs provided ( only "send_goal failed" )
• ros2 daemon needs to restart to see the bt_navigator node again ( after the bt_navigator node is re-launched after it crashes ).

Feature request

• See PR BtActionNode does not crash bt_navigator upon timing out when calling action server #2377