On stepdown, the stepdown thread holds the RSTL and POS mutex lock and calls the instance interrupt method which in-turn tries to acquire instance mutex lock. On the other hand, the instance state machinery thread can try to acquire POS mutex holding instance mutex lock. This resulted in deadlocks. (See SERVER-60872 for historical deadlock scenarios). SERVER-60872 will address those deadlock bugs using quick short-term fix. As part of this ticket, we should come up with a long-term solution.

 EDIT: please refer this deadlock case as well