Model always reloads from SSD, RAM not fully utilzed

### Custom Node Testing

- [x] I have tried disabling custom nodes and the issue persists (see [how to disable custom nodes](https://docs.comfy.org/troubleshooting/custom-node-issues#step-1%3A-test-with-all-custom-nodes-disabled) if you need help)

### Expected Behavior

Using commit:
https://github.com/Comfy-Org/ComfyUI/commit/50e5270b86765bac2da70248d61050abba72b19f

```
32GB RAM
12GB VRAM

LTX2.3, int8
Gemma 3 12b, int8

comfy launch -- --listen 0.0.0.0 --reserve-vram=0.05
```
I noticed that on 2nd and 3rd run, models are always loaded from SSD instead of from RAM, hence slower gen times.
My RAM is also not fully utilized, only at 75-80%.

### Actual Behavior

.

### Steps to Reproduce

.

### Debug Logs

```powershell
.
```

### Other

_No response_

Start by reproducing the second- and third-run behavior with commit 50e5270b86765bac2da70248d61050abba72b19f and the provided comfy launch command. Compare model loading, RAM usage, and SSD activity across runs; done means repeated runs reuse RAM where expected and no longer reload models from SSD unnecessarily.

Written by the indexing model from the issue text.