This addresses the bug Mark fixed in a different way. The BUCKET_SIZE macro was intended to efficiently pack larger buckets. I don't believe I had any buckets as small as we have today when I wrote it as I did not anticipate high traffic 4k page allocation from UMA. I also think it's a little silly to have a bucket with one pointer in it. That's a lot of machinery for a single pointer.
This change will slightly increase memory consumption. I can back it down to the existing limits but I like this better.