I think an open system with easy to add 3rd party stores would be better. The official stores could focus on super high trust applications from huge brands and let the market find a solution to the bottom end. However, the current system is about maintaining outsized control rather than providing a good product to consumers.
As soon as you give up the idea of preventing people from distributing malicious software, and they're not even doing a good job of it right now, you can let competition in a curation market solve the problem for you. I'd way rather have a system where I can get recommendations from someone that's an expert in an area. Ex: Like JonnyGURU is for power supply recommendations, but for software / extensions.
If you extend that concept to the mobile app stores, a system where someone from my city could run a store for local businesses would be significantly better for users and developers than what we have now. For developers it would be amazing to go to a local business, show some local ID, and get a signing certificate. For users it would be amazing to have a local store where established businesses with ties to the community all have a vested interest in it's quality / trustworthiness. That would be at the lowest end for tiny apps. For anything bigger, someone could build a brand / reputation around curation. For example, think of something like a specialized password manager extension store.
When it comes to Google I think there are two problems that prevent them from building a better system. First, they're arrogant and think users are too stupid to control their own devices. Second, their search has devolved to be an atrocious garbage pit of paid content that's optimized for SEO. It's a cyclic dependency where Google's failure makes it difficult for users to make good choices. Google interprets that as the users being dumb and makes the system even more complicated / less effective by adding more ML and automation.
That also probably plays a role in the reluctance to open up some of the current systems. The attempts at scaling with automation and ML are such failures (everywhere) the only way to make them look half reasonable is to ensure no one else can build a competing system.