When I looked in the logs I could see that they hit the API with a lot of requests in short succession that generated 403 responses, so we said that they need to look at how they authorise. After a while they returned and claimed our API was broken. Eventually I offered to look at their code and they were like 'yeah sure but your silly little PHP dev won't understand our perfect C# application'.
So I looked at it and it was a mess of auto-concurrent, nested, for-loops. If they had gotten any data out it would have been an explosion of tens of thousands of requests within a few seconds. They also didn't understand the Bearer-scheme and just threw the string they got from the auth-endpoint straight into the Authorization-header without any prefix. Maybe we should have answered with 400 instead of 403, but yeah, that would have been a breaking change and we didn't want or have time to do a new API version because of this.
Anyway, their tech-manager got really mad that I found the issue within an hour that they had struggled with for months, and also had mentioned that the API-adapter was designed for DoS rather than a polite API-consumer and maybe they should rewrite it to be less greedy and maybe also use ranges instead of crapping out a new request per row in in a response and stitching it together again on their end.
A few weeks later they got it running and it was brutal, but our machines could take the load so we didn't think more about it. Later I heard they got performance issues on their end and had to do as I had suggested anyway.
Be polite and pay attention to detail when you integrate with protocols and API:s. At best you'll be a nuisance if you don't, but many will just block you permanently.