When testing applications that are driven by some external application, the runtime may erroneously detect a "stall" because no tasks are ready to process (but will eventually be once the external application updates a channel or message).

The workaround for this is trying to time some thread sleep before hitting a runtime stall (to "guess" how long a response might take) but this is suboptimal. We should instead offer an option to not exit when there is nothing to do and instead just wait on all pending tasks.