#2167 v0.1.43.1 and v0.1.43 are slower than v0.1.42, and outputs that are not truncated in 0.1.42 are truncated before completion in v0.1.43 and v0.1.43.1, with no errors in the logs

Descrição

Both v0.1.43.1 and v0.1.43 are slower than v0.1.42. In addition, for some inputs whose outputs are not truncated in v0.1.42, the output is truncated before completion in v0.1.43 and v0.1.43.1. The logs show no issues and are identical to the logs from a complete output.

For these truncated cases, the output always stops right after the following part is generated:
python

response_blocks.append(
                f"""<|im_start|>
function: {func_name}
arguments: {json.dumps(args, ensure_ascii=False)}
result: {result}

Environment:
    CPU: Intel Core i9-14900KF
    GPU: NVIDIA RTX 5090
    RAM: 96 GB
    OS: Windows 10
    Model: Qwen3.8-Flash-Next-GSQ-RCO-GGUF\IQ3_S

[strata-iq3_s_v0.1.42.log](https://github.com/user-attachments/files/33356403/strata-iq3_s_v0.1.42.log)
[strata-iq3_s_v0.1.43.1.log](https://github.com/user-attachments/files/33356402/strata-iq3_s_v0.1.43.1.log)