Attempted to recreate suspected segv when loading gid before margo_init(), but unit tests aren't triggering the problem. Probably still helpful to include in test suite, though.
One unrelated fix included for cleaning up ssg internal state after failed init so that subsequent inits work properly. Probably only relevant for unit tests that repeatedly set up and tear down scenarios.
Attempted to recreate suspected segv when loading gid before margo_init(), but unit tests aren't triggering the problem. Probably still helpful to include in test suite, though.
One unrelated fix included for cleaning up ssg internal state after failed init so that subsequent inits work properly. Probably only relevant for unit tests that repeatedly set up and tear down scenarios.