I don't think it's a good idea to jeopardize scientific integrity in the fight against exorbitant subscription fees, that's going way too far. As convenient as Sci-Hub and Libgen are (everyone uses them because they are convenient), you seem to be blissfully unaware of the public library system and the fact that you can order any book or article from any research library in your country. Maybe it's a generational thing, I still remember going to large research library on a bike, ordering 30 articles and books and spending hours at the copying machine. It is possible to work that way, just less convenient. Anyway, the point is that it should be possible to fight expensive subscription fees and at the same time give credit to where it's due.
My point is if you simply don't consult the paid sources in the first place, you don't have an attribution crisis (and then presumably the world gets better and suddenly we have flying cars because 100% of academic research becomes accessible and checkable by all). If we don't read them, we have no need to cite them. Many fields (especially CS) are at the point where you can 100% avoid a trip to the library (or to JSTOR) because the amount of readily available free sources (free as in free for non academics). Even older papers like those by Turing are all available free online and are thus completely fair game to cite without any guilt, and then literally everything new worth citing these days is in arXiv anyway. At least in computing, I guarantee 99% of the time you can find an alternative (and probably better) free citable source for [insert fact here].
When we cite a paywalled paper, we muddy the integrity of our field by saying "sure, you can check up on this, if you pay that is". The downstream effects of everyone doing this is that there are quite simply more inaccuracies out there in current publications. Replication studies literally become more expensive because to replicate 1000 papers, you have to get by 1000 * x paywalls, so we get less replication studies, so the field suffers. If we want accurate research, everyone needs to be able to access every single citation without issue or highway robbery just in case a janitor without JSTOR access finds a flaw in one of your citations' differential equations.
This also leads to the (damaging) centralization of power -- only large institutions can afford to do things like replication studies, and only academics at large institutions can afford to read paywalled citations. It's like we're living in the dark ages. This made more sense when paper and printing were incredibly expensive, like 300 years ago.
In this way, institutions paying for these subscription memberships actually does damage, since they are keeping this model profitable. They are subsidizing the destruction of integrity in fields they care about.
