My logic under removing that comment was that before this MR the use of intermediate array and alerting after the loop was some added local array code, and comment probably served for explicitly justifying these additions. With the MR there is no special case for handling inside or outside spin lock.
There is, though? We do the last wake outside the lock, vectored or not. We'd do the whole wake outside the lock if we could, but we don't know how large it'll have to be.