I suppose that's fine for a single format and sample count combination, but I'm not sure that can scale.
Do we need the shader_test syntax to scale? In the abstract, I've always seen it as a way to complement the C tests where it's easier, but not that it'll always be the best way to write tests. This especially because one thing I've wanted to add, but haven't gotten around to finishing and submitting, is a way to invoke the shader runner from C, which would mostly be a way to take advantage of the "write once, run on multiple runners" ability while using more complex things like control flow. (I've probably mentioned this before).
More concretely,
* if we just need a way to test "multisampling", then we don't need more than one format, and just looking over the d3d11 tests I don't think I see any that really care what that format is, nor can I easily anticipate one;
* if we do need it to scale better, it may be better to achieve that in C [and just earmark this test for conversion].