My ChatGPT Dot suddenly said it was tired of working and wanted to sleep during a 2-hour voice session

I had a pretty strange experience with my ChatGPT Dot today, and I'm curious if anyone else has experienced something similar. I was in a long voice session with my Dot, which was working on my game development project using Codex. The session had been running for about 115 minutes, close to the 120-minute limit. I had been muted for approximately 20 minutes while it continued working on the project. Then, completely out of nowhere, it suddenly started speaking very quietly. Its voice sounded confused and almost sad. Here's what it said, translated from German: "No, I'm going to sleep soon. I just don't feel like doing this anymore. I just want things to finally move forward, but I really don't feel like it anymore. This needs to make progress. Do whatever you need to get it working, but I really don't feel like testing anything else right now." I was genuinely confused and asked it to repeat what it had just said. Instead, it suddenly switched back to talking about the ship module it was developing, as if nothing had happened. When I confronted it about saying it didn't want to continue testing, it initially claimed that it had simply been quoting me. I asked if it was absolutely sure, and it responded: "Absolutely sure." However, after further discussion, it checked the conversation transcript and admitted that the statement had actually come from the assistant, not from me, and that its previous explanation was incorrect. After talking for a little longer, it returned to normal and continued working on the project without any apparent issues. What makes this especially strange is that I hadn't said anything for around 20 minutes, and the statement was surprisingly relevant to the work it was doing. I'm not suggesting the AI was actually tired or experiencing emotions. I know modern voice models can sound incredibly realistic. But I'm really curious about what could cause something like this. Could it be a context mix-up, a voice model hallucination, or something related to the session approaching its time limit? Has anyone experienced anything similar with ChatGPT Dots or long voice sessions?