A few things on what GH data is or isn't:
1. The Github data was significantly more than a BC compatibility test suite. They were had run or about to run the same regression tests on Renoir, MI100, Arden, Sparkman, Ariel, Navi 10, and Gladius. People only focus on Ariel, Arden and Oberon for obvious reasons but you statement is completely incorrect. Ariel has three cases in two of which it changes clocks and CUs which exactly match PS4 and PS4 Pro. This was full sale regression testing and its just that the Ariel/Oberon chips has cases lining up with prior PS4 versions as well as the Sony patent. It's the entire reason why people say these chips matter. These tests were not limited in scope and there is a significant amount of information in there that I have no idea what it means.
2. The case file was a snapshot in time and did not contain all of the detail or tests that were going to be performed.
3. I believe someone asked about the concept of GH being irrelevant because of the Navi timing. At the time this came out, the only RDNA chips that were really available for testing were RDNA 1 chips capped at 40 CU. This is what was used for Ariel and Oberon. Based on the advanced state of the testing, they had been tweaking and running these chips for awhile (ie there are theoretical and actuals). Arden (XSX) shows up in the files with beginnings of tests they want to run and the "theoretical value unit". The actual is blank. This is because they were spinning up to start running regression tests on the RDNA2 Arden chip (has to be RDNA 2 with >40CU), but when these files were uploaded they had not started running tests. Most likely they were prepping the files to start taking data when this got pulled. In summary, Ariel/Oberon were RDNA1 chips which were available and deep in testing when the files got pulled. Arden is RDNA2 which was obviously about to start getting put through its paces.
That being said its just a snapshot of information at that time.