#1519 the engine stopped unexpectedly (exit code 1)

Description

I'm working with Qwen3.8-Flash-Next and in particular with the model qwen3.8-flash-next-unsloth-ud-q4_k_xl. I have an RTX 4090 with 24 GB of VRAM and 128 GB of shared RAM. I'm trying to use strata with VSCode and Kilo Code but the engine stops and restart frequently.

The model is located in a SSD storage.

I have enough free RAM because the model itself is 71 GB of RAM and when I work I just use VSCode, Firefox. So most of the time I have free 40 GB of RAM.

I'm using a custom solution named ai-flow. ai-flow is a lightweight workflow tool for structured AI-assisted software delivery. It keeps planning, tasks, work items, implementation evidence, and independent audits consistent through versioned Markdown/YAML artifacts.
It uses a context-switching approach:
- The cloud model (OpenAI GPT-6 Astra) acts as the architect: it consolidates requirements, decisions, risks, open questions, and decomposes approved work into verifiable tasks.
- The local model (Qwen) acts as the implementer: it receives one bounded work item at a time, changes code, runs the required checks, and records factual evidence.
- An independent audit (OpenAI GPT-5.6 Sol) then reviews the evidence against the acceptance criteria without automatically approving or advancing the task lifecycle.
This separation keeps architectural reasoning, implementation context, and verification context focused and traceable, while reducing the risk that an implementation model silently changes requirements or self-approves its own work.

[strata-unsloth-ud-q4_k_xl.log](https://github.com/user-attachments/files/33195353/strata-unsloth-ud-q4_k_xl.log)