fooocus generates past/wrong image

[dscsqrl]

Hi! I have Focus set up on a Linux server, running through a daemon that activates a Python virtual environment under a special user. Nginx is configured as a proxy. I noticed a strange bug where when I switch the model in the web interface and try to generate an image, and at the same time another person tries to generate an image, my generation hangs, but the second person gets my generated image instead. Nothing happens on my end until I refresh the page. Any ideas?

My focus settings
{
"advanced_mode": true,
"image_number": 1,
"save_metadata_json": true,
"save_metadata_image": true,
"output_format": "jpg",
"seed_random": true,
"same_seed_for_all": false,
"seed": 0,
"styles": ["Default (Slightly Cinematic)"],
"prompt_expansion": true,
"prompt": "",
"negative_prompt": "",
"performance": "Speed",
"custom_steps": 24,
"custom_switch": 0.75,
"img2img_mode": false,
"img2img_start_step": 0.06,
"img2img_denoise": 0.94,
"img2img_scale": 1.0,
"control_lora_canny": false,
"canny_edge_low": 0.2,
"canny_edge_high": 0.8,
"canny_start": 0.0,
"canny_stop": 0.4,
"canny_strength": 0.8,
"canny_model": "control-lora-canny-rank128.safetensors",
"control_lora_depth": false,
"depth_start": 0.0,
"depth_stop": 0.4,
"depth_strength": 0.8,
"depth_model": "control-lora-depth-rank128.safetensors",
"keep_input_names": true,
"revision": false,
"positive_prompt_strength": 1.0,
"negative_prompt_strength": 1.0,
"revision_strength_1": 1.0,
"revision_strength_2": 1.0,
"revision_strength_3": 1.0,
"revision_strength_4": 1.0,
"resolution": "1152�896 (9:7)",
"sampler": "dpmpp_2m_sde_gpu",
"scheduler": "karras",
"cfg": 7.0,
"base_clip_skip": -2,
"refiner_clip_skip": -2,
"sharpness": 2.0,
"base_model": "sd_xl_base_1.0_0.9vae.safetensors",
"refiner_model": "sd_xl_refiner_1.0_0.9vae.safetensors",
"lora_1_model": "sd_xl_offset_example-lora_1.0.safetensors",
"lora_1_weight": 0.5,
"lora_2_model": "None",
"lora_2_weight": 0.5,
"lora_3_model": "None",
"lora_3_weight": 0.5,
"lora_4_model": "None",
"lora_4_weight": 0.5,
"lora_5_model": "None",
"lora_5_weight": 0.5,
"freeu": true,
"freeu_b1": 1.01,
"freeu_b2": 1.02,
"freeu_s1": 0.99,
"freeu_s2": 0.95
}

Even if you run Focus without Nginx, directly on the port, the issue persists.