Thus far these are the failures we've seen:

* not running when we expect it to be running:
   * interpreters.is_running(interp)
   * interpreters.run_string(interp, ...)
   * interpreters.destroy(interp)
* can't find the interpreter even though we expect it to exist
   * interpreters.run_string(interp, ...)
* finds it running when we expect it to not be running
   * interpreters.run_string(interp, ...)

Except for the last one (which might be a separate issue), they all look like they could be explained by the same thing: the subinterpreter stopped (or went away) prematurely.  That could be related to the code in _xxsubinterpretersmodule.c or it could be the cleanup code that makes sure interpreters get cleaned up at the end of tests (e.g. running too soon).  Either way I expect the fix will be in the module code and not the tests.

Regarding "is_running()", notice that it relies almost entirely on "frame->f_executing".  That might not be enough (or maybe the behavior there changed).  That would be worth checking out.

@aeros, feel free too keep investigating.  I'd be glad to help you out.  Otherwise I'll dive into this probably next week.