I have always created shields to go onto top of netduino's for the simple reason that if either fails you can swap the part out, also you could conceivably port your netmf code to arduino, or any of the 'duino foot print dev boards out there.
That said if you want to design your own board then you will need to download the schematics from the downloads section, transcribe them into your CAD package of choice (i.e. design spark, or kicad the free copy of eagle doesn't support enough PCB layers or pads), remove the parts you don't want then route the board. Something to be aware of is that AFAIK the ND boards are 4 layered boards which add a lot to the cost of board production and complexity of routing. Also Secret Labs have spent a lot of time designing the boards to minimize EMR which may or may not prohibit you from selling your custom board depending where you live in the world (EMR testing and compliance certification can cost thousands). So in short yes its possible to integrate all into a single board but its a frankly huge undertaking and would be much easier to design a shield instead instead of reinventing the wheel.
as for small production runs you could try http://oshpark.com/ or http://http://www.se...tml?cPath=70_71 for your board fabrication
as long as u use the same mcu and keep the pin allocation the same then firmware should run as is, and as for flashing your program you could use mfdeploy to deploy to each board once fabricated and assembled. It looks like it would be possible to modify mfdeploy to support multiple (parallel) deployments to several devices at once, but AFAIK no one has made such a modification.
Update
I also forgot to mention that for small runs i.e. < 1000 you probably wont be able to source all the components for the ND and get the board fabricated for less than the cost of a ND.
Also factor in at least 2 prototype runs of your custom design as guaranteed your first one will have some issues (parts to close, incorrect routing), plus the debugging and verification needed to ensure that your ND clone is 100% compatible etc.
All in all if you total up the time spent not working on your application but spent on cloning the ND and getting it working as well as the $40 dollar board you would be looking at several hundred hours of work, on top of your project.
So lets say you could do everything in 350 hours, and assume you would charge $20 an hour if you were doing this work as a contractor you would be looking at a cost of $7000 to develop the device plus the cost of the components (approx $35 per board just for the ND components and pcb as i assume you wouldnt be buying in volume) , then to break even (no profit) the cheapest you could sell the device for is $73.50 assuming you have 100 made plus the cost of the components for your project.
Where as if you were to just build a shield then its just the cost of your components and board run and time spent getting your project up to production standards.
Annother factor to consider is getting the right tools, at the very least i would say you want a logic analyser (cheap ones start at the $50 mark), but a scope would be very useful possibly essential based on your project needs ($350-how much 
), plus surface mount soldering gear, say a hakko fx series solder station ($100+), tips ($10 each), lead free solder, possibly a hot air rework station), so factor in $1000 to get production quality tools as well.
That said you can get away with whats to hand but if you are considering production you have to be able to achieve a pro level of quality and service and cheap kit wont help that.
Sorry if this seems very negative I just cant see any commercial sense in cloning the ND just to have a single board solution, that said designing a shield is much easier and cheaper.
Hopefully this hasn't put you off your idea and i look forward to seeing what you come up with 
Nak.
Edited by nakchak, 25 July 2013 - 11:58 AM.
