Maybe you don't use it very often? For months before the recent chaos it was very reliable. The fact that it's working today is only because I fixed it, Twitter didn't change anything.
It's disabled on nitter.net to reduce load while this new solution is being tested.
Check the instance list for updated instances, most of them have RSS enabled: https://github.com/zedeus/nitter/wiki/Instances
This isn't actually true, the "Instance has been rate limited" error was a false flag. It turned out to be a bug in the API token parser, found and fixed by a contributor. Nitter instances have yet to run into a real rate limit.
Inspired by Twint I created an alternative web client that relies almost entirely on web scraping. The parser is more complete than Twint's due to the different goals, supporting things like polls and videos properly. It's still work in progress but very usable with features like RSS and convenient profile search.
https://github.com/zedeus/nitter
It doesn't take much to run Nitter really, the nitter.net instance is currently sitting at 310 MB of memory and the CPU requirements aren't too bad. It's running off my dedicated home server with fairly low specs (i3-8100) tunnelled through a cheap VPS to bypass firewall restrictions. It hasn't crashed or (to my knowledge) experienced severe slowdowns despite the huge traffic increase today. Invidious's story is a bit different as the main instance load balances between 10 VPS instances.
Regarding usage stats I don't have any accurate numbers, but the database that stores preferences has 30k rows.
It essentially emulates browser requests, "unofficial API" is just a fancy term for web scraping + using required parts of the public API where guest tokens work (ie. fetching video info, cards and polls).
Yes, but I do have some far future plans about utilizing the "Login with Twitter" authentication method to let people use the API on their behalf, which would allow making tweets etc.