I’m thinking of this as a lightweight (single node) alternative in the same way duckdb is focused on data that can be processed by a single node.
I think / hope sqlflow will be a viable/lightweight/cost effective/easy to operate alternative to flink when working with small-medium sized data (on the orders of <10,000 messages / second)
The system is very fundamentally not designed to support such an approach (much as I agree that it would produce a better result).
The way to break up companies is to win antitrust lawsuits against them. Except in cases where the companies are actually tied together in some meaningful way (ie, they're not fully separate companies in practice), there's no way to link such suits to each other; each one has to stand or fall on its own merits, and on its own timetable.
I think it would be hard to argue that Google, Apple, Microsoft, etc are actually arms of the same company. Thus, each of these Big Tech antitrust suits needs to happen separately, and some of the remedies might end up combining in counterproductive ways, because they are generally not able to consider each other's situations (the remedies have to be based on the facts of the individual case, AIUI).
It sucks, but it's the only antitrust system we've got right now.