map server capabilities [...] due to a priority change
Good questions. Just my two cents on the topic...
I think it makes sense that the biggest interest in 9 arises from those things which you can't get elsewhere. If you already have 8, or ESRI, or whatever, and it does what you want there's no need to buy and learn something new. Sure, if doing what you can do in Arc is somewhat easier or faster in 9 that will drive interest. But if you can only do what you want in 9, and that task is mission critical, well, that's something you can't ignore, and it really drives interest. :-)
As a rough simplification, the unique aspects of 9 are mainly about working with data that is big enough that 8 or Arc are too slow, and mainly about working with DBMS connections and other data access that you just can't do in either 8 or Arc. Historically, people and organizations that have dived into 9 need to work with a million vectors today, and in a way they can't realistically do with alternatives. I think that as part of doing that work they naturally are most likely to speak up about features and capabilities they need which are directly involved in that specific work.
That's different than presentation and cartography, which (after you crunch bigger data) you can do in 8 or Arc or many other packages. Map serving is like that, in that there are many options. If you have 9 you can load PostgreSQL with the results of your data crunching and use free map serving software to publish those results to the web. It may not be as convenient as it would be to have 9 automatically be a map server too, but at least there are alternatives.
That's my theory why there seems to be more interest in filling in holes in things like vector editing. If you're editing a couple of million objects, you need the right capabilities in the tool you are using, as there's not an option to export back out to ESRI and then wait forever. You want to do all you must do in 9. Same with raster extensions and tools, where if you want to compute something in rasters you really want to take advantage of 9, get more capabilities as fast functions, then added for ease of use as transform templates and so on.
That's not going to stop 9 from coming out with server capability. But I think it explains the current priority in the near term on filling in missing parts of editing tools and many other "small" things, some ongoing work with rasters and so on.
I also think that as 9 expands there are more and more people who are choosing it not because they don't have any other choice given the data sizes they're working with, but because they like the flexibility of it, they like the quality, and as more and more tools are being added they don't want to have to drop out of 9 into 8 or back to Arc or some other tool to do something they want in cartography or presentation. So that's a rationale for expanding capabilities in legends and layouts.
One more thing on servers: there are many common issues between things like distributing processes between multiple machines, enabling cloud-resident operation of worker processes, and enabling server functionality of Manifold as a web server publishing to clients, be those clients browsers or Manifold client sessions. The easiest thing in the world is to do a one-of-a-kind web server, like IMS in 8 or mapserver or other classic GIS web servers. People often take one-off approaches just to get the thing out. But then they find themselves doing a different one-off approach just to get out client-server ability between an instance of Manifold as a client on one machine and Manifold as a server on a different machine, and yet another one-off approach for cloud and yet another for distributed processing. All that adds up to inject orders of magnitude more complexity into the thing than doing a unified approach (when possible if in truth some server manifestations really are just different sides of the same coin), and such extra complexity very rapidly builds up to where software cannot be maintained or extended or built to crashproof quality.
So there is a lot to be said for not pushing on the map server right now, but perhaps waiting a couple of months on it until the other server work is further along so that a focus on servers can be done in an integrated, maintainable, flexible way. In the meantime, effort invested into steadily filling holes in vector editing, adding raster capabilities, adding some more good things in layouts, introducing legends, and so on, is effort in a good cause, helping many users.