We don't need to issue requests if the opCtx is already dead. We should check the opCtx before incurring significant costs. We already respond to interrupts during request, but we can save more effort if we bail out earlier.