Feature request

As discussed here: #1540. By design, and in the words of @wjwwood , by default it is NOT possible to "wait on asynchronous things within a ROS callback safely", see #1533
The issue (or more accurately behavior by design) is well know.
However with the current implementation of simulation time which listen to the /clock topic, this behavior breaks the symmetry between the usage of real time (system time internally accessed without the need of subscribing to a topic, and hence always updated) and simulation time (time received on the /clock topic and hence frozen in a callback).
I believe the symmetry should be maintain in order to not draw false conclusions about the behavior of a code developed in simulation VS real time, and thus independently from good/bad code practice.

Implementation considerations

Maybe insulation the /clock subscriber callback on a specific thread ?