This is bad, because the names are supposed to be used as identifiers, but it's not clear that this data is constant.

I agree that looks bad. Is there a WPT issue for this already?

Comment on attachment 8972159 [details] disable.patch LGTM

For the record, the expected output is included in the message so you can tell what the output is. The attachment says the test fails, but it seems the last bit is left off, but it would have said something like "element-wise tolerance of {"absoluteThreshold":0.000027657,"relativeThreshold":0}" So the absolute difference between the actual and expected results exceeds 2.7657e-5. There would also be more text specifying what the actual max absolute (and relative) difference was.