phiber You're right, researching more into this, and it seems the resolution to aspect ratio is the issue here.
As you said, to be native 16K resolution without any cropping or upscaling and suitable for the standard side by side equirectangular 180° VR Video, we need 2:1 aspect ratio which is not achievable currently according to the AV1 latest 6.3 level. Hopefully they will address this in the next/ future levels.
As for the AV1 Tile width, I was talking from encoding standpoint of how we can achieve AV1 16K in the most reliable and fastest way with software encoders like SVT-AV1.
Because, currently, as of writing this, the most reliable and fastest way to encode and decode AV1 16K video is by using parallelism, mostly because of the current software encoders and hardware decoders limitations.
Since the Stereoscopic video sides are logically separate, we could also encode/ decode them separately. And AV1 pretty much supports this and here were the AV1 Tiles comes to play when encoding 16K resolution. This way it can be "fast" to encode and "light" to decode thus smooth to playback.
Because AV1 tiles are not fixed, but "flexible" which means they can be encoded in uniform and non-uniform tile spacing, and can be grouped into tile groups and each group can be encoded/ decoded independently.
This is the reason why RTX 5080/ 5090 have 2 Decoders, mostly to deal with "Multiview Videos" in very high resolutions like MV-HEVC for example:
https://blogs.nvidia.com/blog/generative-ai-studio-ces-geforce-rtx-50-series/
Although AV1 does not mention the same approach officially, but it actually can be done through SVT-AV1 and FFmpeg according to some 16K AV1 encoders as I read.
Personally, I can't confirm any of this, as I never encoded videos higher than 8192x4096 @60FPS @10-bit @400Mbps CBR for both AV1 and HEVC. Because that's the maximum of what my PC can encode easily and my RTX 4080 GPU can decode and play smoothly.
The VR 16K resolution videos are pretty much still undiscovered territory and most of us couldn't explored yet. When it comes to any VR video resolution higher than 8K, I'm still learning myself.
Now, that we'll finally have 16K VR Cameras like "Blackmagic URSA Cine Immersive" which also uses dual 8K sensors to achieve this ironically.
This will trigger everyone in various industries to invest, develop and advance the VR 16K @60FPS encoding and decoding and in both software and hardware for sure. Which will make the encoding/ decoding process less complex for the user and less taxing on the hardware.
Best Regards.