Is there an simple description of the pageview API available anywhere already?
The API is deployed publicly, but we don't want to spread the news yet because the data is not fully loaded into it (there are many Terabytes). However, the spec for it is pretty easy to read: https://github.com/wikimedia/restbase/blob/master/mods/pageviews.yaml (the public endpoint is a bit different, but just to get the idea of the routes available for now)
Specifically, how far back will the API access historical data?
Right now, we're going to load data back to May of this year. That's as far back as we have high quality data. We are collecting use cases for people who need the other data, available in pagecounts-raw and/or pagecounts-all-sites. So let us know if you need that and what you use it for, it'll help.
And, maybe even more importantly, will either pagecounts-raw, pagecounts-all-sites or both of those datasets, be discontinued in the near future due to the (presumably) more powerful pageview API?
No plans to discontinue those right now, but it is getting very confusing to understand the differences between the datasets. Even if we do discontinue or merge some of those datasets, we'll talk about it on this list first and get opinions. And we'll always have some type of static file dump for people who need to do bulk work, we just have to figure out which type(s) of data to keep and which to remove to simplify everyone's life :)