Generating large images can take a lot of time, and at a point you will most likely run out of memory.
Because most models are trained on images that are 512x512 (sometimes 768x768) the AI has a hard time dealing with large images and long aspect ratios.
Caused by a resolution that's too large
To get around this you can generate an image at a smaller resolution and use an AI assisted upscaler to make it bigger.
Going too large will cause details to be lost with regular upscalers. But these can be easily added back in with inpainting. (Inpaint)
The other option is to use Highres Fix, which will generate a smaller image then automatically upscale and add details back in.
This is much more resource intense than regular upscalers, but can offer cleaner results.