For me, I think it is related to VPN: I work from home using a VPN to my company. I installed a linuxserver/webtop on my NAS. This one does not use the my company’s network and chatGPT works just fine with it. My personal laptop, that doesn’t go to my company’s network, doesn’t have any issue as well.
So my assumption is that openai is probably (very wrongly) protecting itself from IP adresses from which too many requests are issued, without taking account company networks and VPN, which is very bad.
But, still, I may be proved wrong.