Currently, internal operations rely on calling setSystemOperationKillable() to allow themselves to be killed by the RstlKillOpThread on step up/down. This design is error-prone and has resulted in bugs like SERVER-40594 and SERVER-43174.

Instead, we should somehow make the RstlKillOpThread more intelligent on what operations to kill without relying on clients setting the flag. I think we should always kill operations that have intention to acquire strong locks (Mode X/S) because they could be blocked on prepared transactions while holding onto RSTL, preventing step up/down from proceeding.