> For that reason, I've only been looking at those that scored 10 (best possible) on Appleby's SMHasher[1] test suite, which is used by everyone who does recognized work in this field.

So it seems that this SMHasher test suite doesn't catch the problem that we're seeing with negative integers.

> I'm concerned that I've been putting way too much weight on "the new" tuple test. [...] that's a minuscule region of the problem space.

I'll admit that it's a miniscule region of the problem space. However, we ideally want a hash that works well for all kinds of inputs. If the hash function is good, it shouldn't be possible to write a hash collision test function which has a significantly higher chance of failing than random chance.