Surprisingly I did not find a question for ROS2 to this. So here we go:

Use case: I want to refactor the emergency stop functionality in our setup.

Currently each node has a "~/emergency" trigger service expanding to e.g. "my_node/emergency". This implies that the caller node will need to have an instance per node to be called.

I tested naming the (all the) service simply "/emergency". Calling it from a single client triggers all the servers. Only the response of the first (?) gets through, but that its fine.

Is this some expected and non-abusive behavior or is this more a current feature that may disappear?

The obvious alternative is to switch to a a topic-pub yet changing the legacy code from "~/emergency" to "/emergency" is easier than rewriting each server as a topic-subscriber. but just in case, if there is something in wrong with pub-sub approach as well, please advise.

All the code is in Python. Managed lifecycle nodes are therefore ASAIK yet out of reach.

Thanks!