The spec says this:

Textures in block-based compressed GPUTextureFormats always have a mipmap level 0 whose [[descriptor]].size is a multiple of the texel block size, but the lower mipmap levels might not be multiples of the texel block size and can have paddings.

But this is not a restriction of any of the major APIs that are targets (D3D12, Metal, Vulkan), that I can find. It is a restriction in D3D11, but this restriction seems to have happened by accident. Because mipmaps can already not be texel-aligned, hardware basically needs to support this anyway.

My sleuthing has only lead me to #623 which added the text, and #363 which is the discussion that proposed the restriction, which references a D3D10 manual page.

I would propose dropping this text from the spec.