Since "20K+ users can clear out low-quality reviews faster than normal users, as it takes only 3 votes (vs 6 recommendations) to delete posts.", I would like to know how many 20kers actually use delete votes on the queue and how many posts results deleted by the action of 3 20kers users in the queue, and possibly how many 20kers votes result lost due the post being deleted by 6 normal "recommend deletion" actions.
With no data offered by Stack Exchange so far, I tried using science. The query "gaps in review task Ids" provides potential candidates for LQ review deletions. After some clicking I obtained a sample of 20 LQ reviews that resulted in deletion of answer, from two sites: Stack Overflow and Mathematics. Invalidated reviews (e.g., post deleted by a mod) were excluded.
Sample of 20 LQ reviews from Nov. 12-13. Data in the format [Delete votes / Total of Delete and Recommend Deletion]
Out of 120 Del&recDel votes, 20 were Delete votes. None of them made a difference in the process: the delete votes had the same effect as if they were 2000-rep user's Recommend Deletion votes.
Sample of 20 LQ reviews from Nov. 13-15. Format as above.
Out of 120 Del&recDel votes, 29 were Delete votes. On 3 occasions, three delete votes were cast, resulting in 20K-deletion instead of review-deletion. However, even on those occasions the total of Delete+RecDelete was 6, which means that none of the 29 delete votes sped up the process in any way.
Assuming the above sample is representative, the LQ review queue wastes virtually all of the delete votes of 20K users that are cast within the review.
Replace "Delete" with "Recommend Deletion" for all users. This will save the allotted delete votes of 20K users, leaving more of them for deletion of questions (and of answers not found in the review).