It's a different question. How fast do you need to move said mass?
If you use a ballscrew axis for maximum transfer of rotary to linear motion, appropriately size and gear the motor, a compromise between speed and brute force can be achieved.

For example, I have a giant 3D printer I built for Maker Faire NYC 2014 with a 4 foot bed of glass and 4 leadscrews powered with a single large NEMA23 motor and a double belt reduction. The bed can move 70 pounds at a fairly brisk pace and one configuration was a 2 amp setup. It was ball screws, running an 8:1 reduction from the motor, and the motor was in full step (non microstepping) but that was mainly due to an 8 bit control board max step count math overflow if I used higher than 500 steps/mm. Mechanically and electrically, as long as I stayed at a reasonable speed for the inductance of the motor, and that in turn with the pitch and gearing determined the final Z feedrate.

I will caution that in general, what you want (a mill/3D printer combo) rarely is optimized for both.
Mill requires force and slow speeds
3D printing is desirable to triple or more the feedrates and really wants way less mass to minimize positioning artifacts.
It's not the electronics, it's the mechanics.

I know my list is longer than 5, but wanted to share my thoughts. I tried to put in order of priority. As always, safety and performance need to happen first so IMO, the heater control is the top of the list. Next is interface security. I think in the industry this is far overlooked. To have this project come out with proper security suitable for corporate networks is key if you want me to promote web based and networking printer controllers to the larger community. Again, this is where you can trump the competition hands down. We already have decent performance and features. I'm all about improving them and love pushing the edge, but I want to see a product that is done right, and key is understanding corporate networks and school networks for wide adoption.

C. Predictive temperature control. This will replace PID. Probably the main benefit is that it will have a simple and fast auto tune procedure.
Q. Better web interface security, including logon IDs and passwords that are not sent over WiFi in clear text even when using an unsecured WiFi network, and passwords that cannot be retrieved from the printer via the web interface. IMO this is essential if you want to expose your printer to the internet by opening a port on your router, or you want to use the Duet WiFi on an unsecured wireless network.
H. Grid-based bed compensation, for dealing with beds or printing planes that are not flat.
O. Support for restore points. This would allow the print to be paused, a restore point created, and the printer shut down - either automatically because of power failure (although a UPS or SLA battery would be needed to power the printer for a few seconds), or manually. Then the printer could be re-started and the print continued from the same point. Only practical for 3D printers that can be re-homed when there is a print on the bed, or for which the motors can be relied upon to retain their positions if they are shut down at an appropriate full step position.
B. Multi-threaded web server, capable of supporting several concurrent connections even when one of them is uploading a file. May also increase the speed of file uploading a little.
D. Support for PanelOne and similar 20x4 text displays, not including the SD card socket - see next item.
E. Support for an external SD card socket, such as the one on PanelOne and the one usually present on displays used with PanelDue. The file upload speed to this card will be much lower than to the built-in SDHC card, so the main use will be if you want to write files to an SD card on a PC and then move it to the printer.
F. Support for multiple independent X carriages. The common case is two heads moving on the same rods or rails, which park themselves at opposite ends when not in use.
I. Faster microstepping, without going to dynamically-varying microstepping. Perhaps a 50% speed improvement will be possible.

P. Control over which access point the Duet WiFi connects to, via the USB port, and perhaps from PanelDue as well.

Things I'd rather not see effort put into based on what limited benefit to problem ratio they provide.
J. Dynamically-varying microstepping, allowing true 256x microstepping at low speeds, with microstepping automatically reduced for faster moves.
I know what that means at the bare hardware level here and mode switching while not losing position seems a horrible idea based on everything I know. Even if possible, the list of how it can go wrong is just not worth it IMO.

I also don't think higher than 2 amp stepper on board makes sense. The EMI, RFI, current surges, and heat on a control board just seems like a battle. If you want that much current, the drivers need heatsinking and separate power bus- AKA External drivers.
If you want to be slick, one feature I would support is breaking out step, direction, and enable solder pad throughhole pin locations. Yes, I know you have an expansion header, I want to avoid remapping and other issues. I hate to even say it, something Smoothieboard has that this doesn't.

Another comment I saw was potentially telnet. Please, use SSH or secure protocol with ability to change the default password- going so far as to force the user to change it. We need to guide users into security. Don't allow any anonymous service (FTP,SCP, Telnet, or other) as a control or access method. HTTPS web with signed certificates. Enforce strong authentication. You can view in anonymous mode, but not control. But also give the user the option to change that.

I ordered the optional thermocouple breakout and just wanted to share notes that ended up in the Delta google group so they are here for Beta users. Thanks to David for posting the information so the real credit goes to him.

David Crocker
When you connect your thermocouple, if it has the usual white and green wires then the white one goes into the left hand terminal as you look into the terminal block from the wires. The thermocouple terminals labelled 1 are channel 100, so you need M305 P1 X100 in config.g to associate that thermocouple with heater 1, in place of the standard M305 P1 command.

I'll try to post some pictures and a larger guide on this soon. For me, I think additional information is also good to share like theory of thermocouple operation, key install and wiring points, special emphasis on understanding cold junction details, as well as the super critical thermal connection but electrical isolation required at the sensing end.
Far too often we see beginner users make massive mistakes and blow up entire boards and systems because of lack of understanding of these key issues.

Also, the knowledge that the typical thermocouple requires no calibration if implemented correctly. Compared to a thermistor, you aren't worried about BETA values and other issues and tables. A K type thermocouple can be as accurate as 3C plus or minus between any 2 printers with zero calibration by the user. But it means following the rules.