Apps loading and executing additional code at runtime (as opposed to e.g. gating behavior using feature flags fetched from the backend) just doesn't sound like a great idea, even if it might technically pass the various app stores' review guidelines.
I do think code push is hard to get right. I think that's one of the values hopefully Shorebird can provide, by offering an easy-to-use and easy-to-get-right solution that's accessible to teams who use Flutter, but don't have an extra 10 engineers to write their own code push solution.
I feel very aligned with the stores in that what I want are great apps for users. Having users stuck on buggy/broken apps is a pain not just for companies but also for the platforms and most importantly the users (including from a security perspective).
But should we then maybe just stop the pretense of "app stores control every executable byte of code in the apps they distribute" in favor of "app stores do content moderation at the business level; what apps technically do is up to their developers"?
Maybe we could even have both models (immutable, signed bundles vs. very dynamic entry points ad-hoc fetching parts of their logic), with app stores indicating which one an app purports to follow.