From #351:

Nitter uses a combination of public and private Twitter APIs, using the credentials they use for their own web client, and guest tokens. Nitter instances do not talk to each other, and no html scraping is done. I hope that answers your questions.

The “rate limited” error messages you may see on some Nitter instances is due to a bug, they are not actually rate limited. The only rate limit that can affect a Nitter instance is IP-based, but it requires a lot of requests.

So weird that Twitter is ok with these stuff being “exploited” by third-party mirros. It feels like nitter is walking on thin ice. Hopefully twitter’s not going to act on this

Anyway thanks for the insight