Models from Uncontrolled Reiteration

Fix models that reiterate an uncontrolled or unnecessary number of times. I recently noticed that some models, especially the latest Gemini "thinking" model, will constantly reiterate their output again and again. A simple user prompt can make the model stack up many repeated outputs, often 30-50 times. Especially when unattended, where the user can’t stop the output, it will drain a significant number of tokens. The attached image shows a single user prompt that resulted in 38 outputs from the model.