Assets packaged as assetbundle should be
significantly faster to load. Also it uses less GPU memory, if you choose a hardware compressed texture format, e.g. like DXT. Unity handles that for you, but you have to choose a matching format.
If you use a regular image file instead, since a graphics card doesn't know what a JPG or PNG is, it has to be decompressed on a single core by the CPU and is then stored uncompressed in GPU memory. That takes time. And an uncompressed 8K image is like 96MB, at least. (By the way there is no point in 16K skies by the way...as maximum texture size Unity can handle is 8192x8192)
Hardware compressed textures need more space on disk, but less memory at runtime (like 1/6) and can also render much faster because of that. Slow memory access is often a reason for performance limitation.
I have not measured anything with VaM, but I would expect the gain in scene loading performance will likely be comparable to the difference between loading regular sounds and assetbundle sounds, which is like 50-100x faster. Of course you would usually only have single sky, not like hundred you have with sounds, so probably takes only a second or two to load, so you may not notice much
When you store multiple skies into one assetbundle, make sure to use chunk-based LZ4 compression or uncompressed, with LZMA compression performance will tank as the whole bundle needs to be extracted. If you can affort the disk space, I recommend uncompressed...because the texture is compressed already anyway, you don't gain that much from compressing a second time, just loosing time when loading it. (not to mention that a VAR would try to compress a third time
)