[ketzu]

I agree with the general point of your comment, but my personal experience with the usefulness of reviews is not so great.

Nearly all reviews were superficial, not a single review ever found any technical flaw, even when some were discovered later. The acceptance recommendations seem to be governed by the overall impression the reviewer had from the paper and not by any factual points they make in the review. This manifested in getting 5 reviews at one conference, where all reviewers made nearly the same points in their review: all about the presentation of the paper. The paper received all of the five recommendations: Accept, weak accept, borderline, weak reject, reject. (The most critical and useful with a small correction of one formula was the "Accept" review.)

Many reviewers also leave the impression they didn't read the paper in full or at all, or ran out of time while for the review, so looked for a reason to reject the paper and didn't look any further.

Note: I am not saying my papers should have been accepted, 2 had serious issues I glad I discovered later! I also had a few very useful reviews to improve the presentation of my paper.