Happens with certain models such as Gemma3:

Repro:

make -j && ./bin/llama-server -hf ggml-org/gemma-3-1b-it-qat-GGUF --jinja

• Start new conv
• Type "Hello"
• Wait for response to finish
• Click regenerate
• Wait for response to finish
• Type "Test"
• Observe error