Your experience matches mine exactly.
I’m a translator and use GPT to assist with my work, so I’m very sensitive to changes in tone and style. Everything was completely normal on the 28th. On the 29th, the responses suddenly shifted, and I’m confident they are coming from 5.4 Thinking, not 5.5. This isn’t just a feeling, my data export confirms it. The default_model_slug is set to gpt-5-5-thinking, but the resolved_model_slug on every single message shows gpt-5-4-thinking (screenshots in my original post).
As for support, my case was escalated to a specialist team over a week ago with no follow-up. After two follow-ups from me, they replied three days ago saying the internal team was reviewing, but there’s been no progress or clarification since. It’s been 11 days now and I still have no idea whether this is a bug, a test, or something else entirely.