Advocate IHi Claire. I am seeing the same issue in both our Premium Capacity (dedicated to our tenant) and non-Premium (shared with other tenants). We definitely have most of our Dataflows in Premium Capacity and that is probably true for the other posts here (though I obviously have no way of confirming that).
As for the Azure Data Lake Storage Gen2. I'm not sure which doc you're looking at it, but maybe related to this option (high-level):
By default, the files related to your dataflows (CSV and JSON) are stored "somewhere" that you can't get to directly. There is an option to change that to your own Azure Storage account which then gives you access to those files. I have that in place for some of my Workspaces but it is not making any difference in terms of the Validation performance.
FTR, I have been a heavy dataflow user for more than two years. I want my work to be easily reusable by me and others... While I've grown accustomed to some slowness at the Validation step (especially if comparing to working inside of PBI Desktop), something changed last week. Now things that were taking 10s of seconds are taking 10s of minutes, sometimes close to an hour. Not OK...
Here's something you might want to try out: Even when things are running normal, if I am building new things, I do it from Power BI Desktop first and then copy it to a Dataflow. If I need to make minor changes after, I do it from the Dataflow editor. If I have more complex changes to make, I copy it from the Dataflow to Power BI Desktop and then back again. The copy also brings in any “upstream” dependences, so it is a very easy to use feature. One catch is if you are copying “back” it will not overwrite and will instead duplicate. The way to address that is by using the Advanced Editor to overwrite the underlying code instead.