Interesting problem.
If you can do metallography, or send out and have it done, you may get lucky and be able to actually see a magnified cross section of the leak path, which may give a better clue as to the failure mechanism, such as weld metal solidification cracking, HAZ hot cracking. But it may be that problem is basically due to the material chemistry, and if you are stuck and cannot change material type or lot of material, you may simply be stuck with an elevated failure rate. This happened to me with Inconel 625 product.
You can of course tighten up everything about your material handling, cleaning, and weld process, and maybe even try changing the process somehow. Sometimes none of this helps.
A couple things to think about.
Hot cracking in some alloys does occur in the heat affected zone.
Leaks can occur from inclusion stringers within the base metal, having nothing to do with the weld. I've seen this in some alloy steels, but kind of doubt that a stainless steel would have stringers to this degree.
Let us know what you find.