It's appropriate for HTML5 to try to be something that people will implement (and, if it's successful, this will be something that the W3C unfortunately never managed). But the Acid tests are intentionally designed to be wishful thinking, rather than reflecting reality. They test a bunch of features chosen to be things that web designers would like to be able to get consistent results out of and currently cannot. And they provide a way for web designers to look at the things that they'd like to use, and compare this against what browsers currently support, so that they can decide how to cope with reality.
The best solution is probably to test, first, that the <video> tag is handled and the browser will, if offered H.263 and Theora, pick one it supports; then to test that, if only offered a single file, it will be able to use it. Of course, there's the question of whether web designers wish everyone could handle Theora or whether web designers wish everyone could handle H.263. I personally wish Theora was what web designers would prefer, but I don't know if that'll be true. I expect the actual answer is that some designers want each.