"On another level, however, it's a disaster for about 99 percent of releases, which stand absolutely no chance of garnering any attention, no matter their quality. The solution: human storefront curation, which Valve has never shown any intention of doing."
I'm not quite sure how this is intended to work: if "human storefront curation" is intended to provide better recommendations it's quite possible that it would be successful, though the automated similarity/people-like-you-bought ones already aren't terrible; but that wouldn't really change the fact that the majority of releases die in obscurity. There are just so many that they cannot all be visible at once; only some relatively more visible than others.
If "human storefront curation" is intended to mean tougher reviews; then isn't the effect the same? Roughly the same games that today languish in obscurity will instead just not get listed.
I don't mean to defend Steam's curation and discovery as the gold standard, there's definitely room for improvement; but I just don't see how even an arbitrarily good curation and discovery mechanism, with downright omniscient understanding of what each buyer wants and what each game delivers, will substantially change the fact that there aren't enough man hours available for 14,500 games/year to get enough attention to keep most of them from selling basically nothing. Especially when so many of them are just bad, or OK-ish but a direct clone of a strictly better game. There probably are some undiscovered gems that are tragically unknown and undersold because they fall into some sort of algorithmic blind spot; but there are also overdiscovered messes that probably deserve more obscurity than they get, so curation improvements would cut both ways.