FYI, people who actually understand how this tech works have explained this quite a few times. Here and on other forums as well. The data in the initial Oberon A0 leak was from an Ariel iGPU profile testlist ran in regression mode.
It's nothing to do with fanboys; it's just speaking to that very likely was actually the case given AMD's naming nomenclature for their chips, and given Ariel came before Oberon as well as the fact Ariel was an RDNA1 chip. So even if Oberon is an RDNA2 chip, as it's ran in Ariel iGPU profile and running its testlists, the results would only be configured for Ariel's specifications, and not necessarily Oberon's.
Considering the first Oberon devkit got sent out a bit after E3 2019, and the claim from insiders that XSX devkits were running behind in schedule compared to PS5, yet an APU die shot of XSX's APU was published online in early January, we can assume from there that XSX's APU was at least mature enough to produce some semi-final silicon that the Xbox team was comfortable enough to show off. That would mean MS had RDNA2 silicon to work with months earlier in the back half of 2019 and, again, if Sony's devkit progress has been ahead for a stretch of the devkit timing (as many have claimed), they would've had RDNA2 silicon to work with dating back to the time the first Oberon silicon was produced, and ran Ariel iGPU profile testlists on Oberon in regression.
This has nothing to do with most wishing PS5 is 9.2TF; just because the Oberon information still holds relevance doesn't mean people referring to it want a "bad scenario" for PS5. It's literally just pertinent data, and it gives us a timeline and some context for PS5 silicon production & testing. However if you better understood the actual test case for Oberon then it'd be easier for you to understand why people like myself believe that the fuller Oberon chip has very likely not yet been tested outside of the Ariel iGPU profile, either. Or at least hasn't had testing data found yet pointing to such.
In other words, we haven't seen the entirety of the Oberon chip, and there have been other revisions since the first one addressing silicon issues. The fact that Oberon has seen testing as late as December 2019 indicates that in all likelihood, it is the PS5 chip, so with that new context we can take comments from people like Matt (which mentioned to disregard it) to most likely refer to the testing data itself not being representative of final specifications, because since the testing data ran was Ariel iGPU profile testlists, that would make Matt's statement accurate. He also said that the data wasn't representative of the final chip (or something to that effect), which is also accurate because Ariel's iGPU profile obviously wouldn't be Oberon's.
So with Oberon as the PS5 chip, it just asks the question if E0 is the final revision, and if either E0 (or an F0) are the final revision, how big is the actual chip. Those are things that will hopefully be found in due time. But in the advent of recent news from AMD, I think for people weighing the options about equally, it actually strengthens the case of Oberon rather than weakening it, because there's enough circumstantial information in this pot to make Oberon testing data and insider claims converge to some point of agreement. The question is just moreso what upper range of insider claims are actually accurate. Does PS5's potential TF performance run from 9.2-10, or 9.2-11, 9.2-12, 9.2-13? We literally don't know.
However, it's fair to keep expectations in check. The higher you set its ceiling you also raise its floor; you have to realize at what point pricing comes into play. And we already have a few BOM estimates from a few reputable sources, so we can then extrapolate what type of chip size fits into those BOM estimates and what it would need to hit your assumed numbers.
We can kind of thank Sony's insistent silence on creating this situation where these multiple sources pretty much equally valid for consideration, because they themselves haven't narrowed much anything down. But given what we've got so far, I see no reason to disregard the Github info or the testing data that's been datamined since then, because agreement with them does not automatically indicate someone also believes or wants the 9.2TF number, either! That's why it's helpful to decouple your biases in analysis, otherwise you draw certain conclusions from people by inferring a certain agenda or angle in their analysis when a neutral read on that analysis will likely show you they are just commenting on what information is there and trying to be relatively neutral themselves.