[I stand corrected, redacting the rest of the comment]
In a small defense of Reddit:
If subs are private, they are still accessible to subscribers. If the moderators don’t moderate the really bad stuff, Reddit has historically replaced them. I would think the same applies here.
It’s hard to argue that people who are not subscribed to a sub are community members. People had plenty of warning to join subs before they went private. Going private alone shouldn’t lead to mod replacement.
> If subs are private, they are still accessible to subscribers.
No, this is not accurate.
A private subreddit is only accessible to "approved users", i.e. those on a list managed by the moderators. Subscribing is completely different; it's a private, user-settable flag whose only significant effect is to make the subreddit's content eligible to show up on your home page.
If you can't view a subreddit's content because it's private and you're not approved, then subscribing makes no difference. In fact, subreddit mods can't even see who's subscribed, nor can they prevent anyone from subscribing, even by banning them.
subscribed members cannot access a private sub, only approved members. Approved member is not quite the same thing as subscription, there are some subs you can't even see without being added, like r/CenturyClub. And being on the approved list (sub visible) doesn't mean it shows up in your home either, you can be added or banned by any sub at any time independent of action from you. It's just literally a pick/ban list for that sub and they can build it off your posts elsewhere etc.
Subscription is about curating your homepage, news subscriptions (ugh), etc. Subscription means you want to see and know about popular stuff in that sub. And you can still view stuff you're not subscribed to, if you're an approved member.
The access levels that are allowable are: approved only, public but only approved can submit links, and public. You can also set access on items individually, and lock comments on those items, etc. So it's also possible to just blank or remove all the old posts with a bot (API!) and leave only a "fuck u" thread that's uncommentable.
At that point reddit probably reverts it and calls it vandalism and puts in new mods and starts banning anybody who's tantruming. Like, be disruptive get banned. You have less freedom here than with a TLD registrar (better ask nicely to get a .edu or .mil or .gov cert issued). Their system, their rules, their call. You already agreed to license your content to them. Shreddit could probably even be undone if they have history etc. And there already is a public dataset already on the internet for like... 10y+ now? how does that work exactly?
Who wins in the argument of GDPR vs users having agreed to a perpetual irrevocable license and several other parties already having used those API access to republish those comments under CC (fh-bigquery:reddit-comments, pushift, etc)? Can you revoke an irrevocable license under GDPR as the originator, and if so how does that leave IP licensing in general, just "fuck it GDPR it, it's mine, I don't want a licensee to use it"? I mean probably all social media licenses include something like that, if it's durable then GDPR is relatively meaningless and nobody has to respond to GDPR if they publish their dataaset/API output (interesting outcome perhaps). If it's not durable it kind of fucks up IP law and subassignment in general, everything was always done by an originating creator(s) at some point, then sublicensed to their employer, etc. Can you just be like "patent license revoked, boss, I did the GDPR?".
The idea of Google/etc being forced to publish their codebases under permissive licenses to avoid random "nah fuck you"'s from departing engineers would be an extremely funni legal outcome though, like that would have massive implications for knowledge/creative work in general. Maybe the same thing would happen to social media providers... maybe that paradoxically makes the incentive to publish fh-bigquery:reddit and pushift stronger, because now it's irrevocable. And that's the minimum ask for 3PA tooling more or less.
On the other hand if the CC sublicense holds in any way, most likely it's legal to say "ok this is the vandalism reversion dataset, it's CC, patch it over the api response inside the client" both for the vendor and the others.
I guess to put it another way: would wikipedia have to edit out every line and commit a user has interacted with, if they filed a GDPR? How do you begin to back that out on a word by word basis? Every redditor knew their words were going to be public (for the definition of that subreddit, and potentially public according to future changes of that subreddit). They put it on "wikipedia" knowing the organization ultimately was driving there.
Reddit chooses not to unwind history and revert. But does Reddit have a license to do so? What about Wikipedia? It seems like in this case the answer may be "probably, if they want to, it's just normally not a problem but we can do it if people get disruptive". People licensed their posts to the community, that's why they're outraged that the community is being torn away from them, it cuts both ways unfortunately. But it might also be legal to reboot Reddit using that data, assuming it's all really CC etc.
In a small defense of Reddit:
If subs are private, they are still accessible to subscribers. If the moderators don’t moderate the really bad stuff, Reddit has historically replaced them. I would think the same applies here.
It’s hard to argue that people who are not subscribed to a sub are community members. People had plenty of warning to join subs before they went private. Going private alone shouldn’t lead to mod replacement.