As someone who actually teaches STEM courses in schools and has to deal with the available infrastructure, the existing $25 Model A (no Ethernet and one USB port) and $35 Model B (Ethernet and two USB ports) is preferable to your suggestion. Building in WiFi would cost at least as much as adding a WiFi dongle (even a couple of million Pi units is paltry compared to the number of WiFi dongles manufactured each year) and would still consume a USB port in any case. A Model A with a WiFi dongle is still less than a Model B and can be remotely accessed using smart phones, tablets, laptops, or anything else with WiFi capability (which does not include most computers in schools). Same thing for BlueTooth, which provides even less value-added.
Our schools, in a relatively well-off area, still only have one desktop or laptop per classroom for teacher use that's on wired Ethernet, a computer lab with about 40 desktop systems on wired Ethernet, and four carts with 20 ~ 30 netbooks each that have to be checked out and moved around between classes in a complex dance to prevent damage/theft (average class size is 35, so one cart per class is never enough for all students). Classrooms are getting WiFi access points, but can't handle more than a few dozen systems' worth of access at one time as the wired network the WAPs are connected to can be as slow as 10 Mbps for an entire school of 1,000 students, teachers, and administrative users. YouBoob viewing is blocked for bandwidth-conservation reasons except in some cases for teachers' wired systems, but even then it's essentially useless during the school day and only useful when material can be downloaded (which most YouBoob videos are not due to default protection) outside class hours.
It's not necessary or even desirable to provide Internet access to Pi boards in the classroom as it's just a source of potential distractions and problems. We typically do as much as possible with local proxy servers and pre-configured local storage to provide content that's been set up in advance to minimize the impact of inevitable network degradation/outages. That also makes it simple to reset systems to a known state with lesson material ready to go, although the Pi's SD cards also make that much simpler than dealing with multitudes of disk drives. We can simulate Internet access by copying relevant server content locally (e.g., Wikipedia, educational sites, news sites, etc.), remapping domain names to IP addresses for DNS, etc. Local networking via Ethernet or WiFi is only used after students have mastered stand-alone system fundamentals such as file system basics, software development principles, GPIO interactions, Arduino programming, etc.
All of the cards seem to point to the next Pi not being on the horizon until probably 2015 at the earliest, and probably only then if the current components are no longer available, e.g., Broadcom no longer manufactures any more of the BCM2835 system-on-a-chip (SoC) devices. Remember that the Pi was developed by volunteers for educational purposes, not by engineers being paid to fast-track a design on a more typical commercially-aggressive, profit-oriented schedule. A next-generation Pi will need to be built using components available at their lowest cost then, so there won't be any custom devices being made given the Pi's continued low price point that is possible in the largest part due to use of last-generation components no longer being used in the latest commercial products. If those components happen to provide things like WiFi and BT at a low enough cost, then they can be used, but increasing cost to get them in is not likely to be compatible with the Foundation's goals.
The best things in life aren't things ... but, a Pi comes pretty darned close!
"Education is not the filling of a pail, but the lighting of a fire." -- W.B. Yeats
In theory, theory & practice are the same - in practice, they aren't!!!