The pipeline test failed again because there were tests in `static-initializers.shader_test` and `arithmetic-int.shader_test` that were retrieving the unexpected hr 0x88760b59.
I had previously added this mapping to E_FAIL in 7/10 because this also happened for the first test in `register-reservations-numeric.shader_test` but now I had to move that check to 6/10.
By the way, this shows a little inconvenience with 6/10: For now, it is hard to catch the behavior of sm1 compilation on all the tests before the pipeline does, because that only happens on Windows in the shader_runner (and I don't want to overload this MR even more including the patches necessary to run SM1 compilation tests on Linux!).
I would probably have to add some more `todo(sm<4)` qualifiers once I rebase this series.