The paper file listing [1] from the PC in question shows predictable [conference][year]-paper[incrementing-number] file name pattern. Rather than "insider help" [2], could this be explained via a "leaky endpoint/api" that got scraped?
I hope I'm missing something.
[1] https://miro.medium.com/max/1400/1*sNC6SuC1v8peYmw6KEcz3g.pn...
[2] Quote: "HotCRP does not show you your own submission even if you are a PC member, so how did the de-anonymized reviews and PC comments for Huixiang’s paper end up in his laptop? An insider help seems likely."