They used to publish a regular torrent of everything, not sure if they still do that? Maybe you could grab yourself a copy for posterity.

It's still possible to get a dump from the Settings page, just with a checkbox that says "if you use this for AI training we might ban you".

You can automate it with browser automation tools.

Perhaps not what you intended, but most of it is stored in various models' weights.

For better or for worse, the contents of Stack Overflow are literally the reason LLM coding has taken over the world.

No, that's been mainly due to RLVR.

Up until a few years ago, SO published full dumps onto the Internet Archive. There has been a community effort which has picked up the torch and is continuing to do the same on a ~monthly cadence.