I discovered a totally ad-hoc special case hacked into SemanticQueryBuilder that reinterpreted value = null to mean value is null.

This is completely contrary to nature, every dialect of SQL I’m familiar with, the JPA spec, and common sense.

There was a comment justifying this on the basis that a couple of tests relied on it, but I looked at those tests and they were testing completely unrelated functionality.

So this is a bug and should be fixed.