About your points:
1. this is totally the point of data mesh. Through domain driven design of the analytical data, only use valuable data.
2. yes it is a paradigm change, that each domain needs to care for its own data, but a similar paradigm change happened already in a lot of companies for e.g. operational topics or BI development. Sure, this won’t happen overnight, but I can see a future where a sophisticated platform for your datamesh simplifies the processes to a certain degree, where the actual data transformations can be handled by the domains (surely a lot of analytics/data engineers will be relocated to these domains). Think of the simplifications which already happened in data engineering. From self hosted on Orem Zookeeper + Hadoop + hive + spark whatever to managed cloud stuff like databricks + dbt. This is already sooo much easier then the past.
3. see 2, this will be a long process, but the actual data transformation process needs to be simplified and central data engineers need to switch to the domains for data mesh to work.
4. this is something only time will tell, but tbh, central governance is also not really working well nowadays
5. as stated also in the data mesh book, stuff like this needs to be established first with data mesh in mind and there will be a lot of improvements in the future
6. I would argue exactly the opposite. So many people do everything to get their hands on valuable data within a company. Just think of all these excel sheets which are flying around in so many departments. If you can establish a really easy way to work with data, many people will love it.