This approach works very differently and is not reverse compatible with the flexbox model in the original post.
The way your code is set up only works if your images are the last element in your file, and there are some additional quirks with the span modifiers. Because of the way Obsidian changed how images are grouped, I don’t see a way that it could be set up to be reverse compatible so that you can use multiple image grids within the flow of text.
I recommend splitting this approach out into its own thread to describe how it works. It might still be a good fit for some users!