Came here as promised 😄 Pixelkiln already gave you a much deeper technical file analysis than I could, so I won't repeat that. His points about consistent canvas sizing, padding and especially the thumbnail are definitely worth looking at.
From my perspective as someone actually building a UI-heavy game, I think the biggest question is consistency as the pack grows.
The four characters work as a starter set, but if you're planning 32 teams, I would establish a very strict template now: same canvas size, same character scale, same ground line, same amount of empty space around every character, and ideally the same visual treatment. Then a developer should be able to swap Portugal for Brazil or any future team without touching their UI layout at all.
I also agree strongly with the thumbnail point. You're selling an avatar pack, but the avatars aren't currently the strongest thing people see when browsing. I'd make 3–4 characters large and immediately readable in the cover. Let the product explain itself before somebody even reads the title.
One more thing from my own recent experience: because you disclose that these are AI-assisted and hand-cleaned, I would make the hand-cleaning and consistency part of the value proposition rather than trying to hide the AI origin. I've received criticism about AI artwork in my own project too, and I've learned that consistency, usability and transparency matter a lot.
Overall I wouldn't rush into producing all 32 teams yet. Fix the technical issues Pixelkiln found, establish one strict production template, improve the store presentation, then make maybe 2–4 more teams using that exact pipeline. If those work without manual adjustment in a real UI, then scale it.
You asked for harsh feedback, so there it is 😄 But I think the concept itself absolutely has a use case.