[zerocrates]

The strategy here is that you have a real condition which isn't index-compatible, so to improve performance, you add on an extra condition which is index-compatible and is... in the ballpark of the condition you want, specifically containing your entire real desired result set, plus some spillover where it's not quite filtering tightly enough. The idea being that the broader condition that uses the index will get done first, then the slow row-scan will happen over a much smaller set of rows: same results but better performance.

But, because the index-compatible condition is too broad, you still need to have the "original" one to get the correct results. If your new, "redundant", condition gets the exact correct results then you're not doing this pattern, you're just replacing a query that can't use an index with one that can, and in that case sure, it doesn't make sense to keep the old one around.