Thanks for the explanation @ebrehault. I suspect what is going on here is that diazo provides two main benefits, and wildcards way of working only needing one of them. I suspect wildcard doesn't reuse a lot of other peoples plugins and when has python developers who create custom tiles and other files system packages. They may also use jbot which requires an experienced developer. We on the other hand use diazo, not just for layout, but to fiddle with html from 3rd party plugins to get what we need for a given them. We do multisite exclusively so never pollute our add-ons with theme specific overrides of html, like you can do if you have one server per site.
The other reason why the diazo approach is still useful is when working with external themes and designers. We recently did a theme that had to get redelivered for various reasons. Each time it cost us almost nothing just reupload that theme overtop. Sometimes a couple of rule adjustments. If you go with dismantling a theme into seperate html parts then you can no longer do that. This only works when you have designers who are very technical and in lots of communication with your plone developers.
That is just our experience. There are probably other ways people theme out there that should be considered.
Do you think its wise to make such a big direction change without wider consultation with people who theme a lot of sites?
It's really important to have the right people in the room when these discussions take place. The way the big boys do things is not the way everyone does things. This is why rationales and wider community outreach is so important.