I think there are two questions we're trying to answer:
- What are portals for?
- (Therefore) What should they contain?
I think portals serve multiple purposes, but their primary purpose is to showcase content from across the project(s) relating to a particular topic. I say "project(s)" because portals have historically showcased information from WikiNews and the Commons as well as Wikipedia itself and arguably could/should include content from Wikiquote, Wikivoyage and other sister projects too. I wouldn't object to "Wikishowcase" being established as a project in its own right and would gladly participate in it. Maybe we should suggest that at Meta and see where it goes. Navigation is clearly a key part of what portals offer; particularly the bringing together of different kinds of content (lists, articles, categories, pictures, news, project tasks, etc) onto one page - including, as I've just said, content from the sister projects. Because portals do that, they're a good way of making visitors aware of what's available and what needs doing. More than once I've found myself editing articles or performing wikignome-like actions in topics I don't normally edit in, or in sister projects I hadn't previously been involved in, due to visiting a portal. So they're a good means of highlighting what needs doing, as well as showing off what has been done. In terms of selected content, I think there are three broad categories of selected articles/biographies etc:
- "Recognised" content - featured/good stuff. Because we want to showcase our best work. This kind of selection could easily be automated using User:JL-Bot - most portals and project already have a page listing recognised content.
- All directly related content - essentially the contents of a related navbox, which of course is the approach TTH took with the controversial mass creation exercise. The mass creation thing itself was perhaps somewhat unwise, but there's nothing wrong in theory with a navbox forming the basis of a portal's selected content.
- Less directly related content - for example, all/most of the wikilinks on the main article about a topic.
The key thing to note about all three of those is that they can all be generated and maintained automatically - so the Phabricator type approach could work well. But perhaps the crucial point is that there seems to be a notion around that portals should require some kind of regular overview by a human being, and that selected content should perhaps be hand-picked using a combination of those. Personally, I don't really see what advantage that has for readers or editors. Manual curation can be fun, but has the potential to introduce WP:POV problems that aren't there with fully automated content, as well as the ongoing need for volunteers to carry out the maintenance. Offering the reader the opportunity to add their own suggestions to the list of selected content can be attractive and might give them a first taste of wiki-editing, but I'm not totally convinced by that. In any case, it would require the portal to be deliberately unfininished. (I read the other day that one of the original "rules" of Wikipedia was "Always leave something for someone else to finish" - I wish I'd made a note of where I found it for reference!) which might or might not be a good thing.