It’s not entirely unexpected but users have adapted by now, added user agents and our servers are back at being at capacity. Adding another server unfortunately has become so expensive that it is not an option. We have to make due with the resources we have.
In the last few weeks there has been a rather high number of requests being rate-limited. I presume that this is due to more people implementing retries when the server returns an error. This is really not helping as it only increases load on an already overloaded server. I have now further narrowed down the limits for banning users from the server. Clients that repeatedly run into a 429 or 406 will be more quickly banned. If you are using Overpass in a script or through auto-reload on a website and run into a 429, make sure to wait at least 30s before sending the next query. For example, when using curl with --retry, make sure to add --retry-delay 30.
I’m also going to ban apps that show a very high rate of 429 returns. Furthermore, I’ve now banned many platforms for fast-deployment of AI-generated apps like lovable.app or netlify.app in the hope of making people aware earlier of the resources they are using.
I hope that at some point we can narrow down the really problematic uses that have caused throughput to go down so dramatically in the last year. In the meantime, please be considerate that Overpass is a shared resource.