@Kalman --- that wont get round the contention, if you use the method shown in the article you can throttle the deletes to avoid locking for too long a time, whilst only perform one select per large batch to get the ids you want.

I do something similar on a 24/7 db, 365 days a year with all files on a single raid 5 partition!?! I use the same approach but with a table variable and I process over 1.5 million deletes per day on a rolling basis.

The cte will keep repeating teh select for the delete and therefore likely cause locking for the duration of the select/delete.