Been contemplating having a go, time is the issue.
1: Don't use copper, the heat load is just not that great and copper is a bugger to machine. A good grade of aluminium should be quite sufficient.
2: Rigidity will be king, so I would avoid peltier cooling (which in any case just increases the total heat load (and temperature differentials)), a big heatsink in a separate compartment from the optics with a blower is IMHO a better approach.
3: Fast axis correction, try to use optics sufficiently large to handle multiple beams, and maybe collimate the slow axis right at the diode then follow with the fast axis anamorphic telescope.
4: You might wish to investigate rod lenses installed across the front of the array.
5: "Design and mounting of prisims and small mirrors in optical instruments" (Paul R. Yoder) has an interesting drawing of a single element anamorphic prism that works by TIR.
6: The diode emitter is a line, use this when you plan the optical layout.
7: If doing a 3 *4 array or such, design the mounting so you can knife edge multiple diodes off one mirror, it means you need two mirror arrays (a 3 and a 4 mirror group) but that is better then 12 individual mirrors.
8: Think carefully about what needs to be adjustable, not everything on the optical bed does and if it doesn't then machining a small alignment step and just epoxying the thing in place is one less thing to slip out of trim. Get a good book on the kinematics of machines, it will provide enlightenment.
9: At this power level, be careful out there, even diffuse reflections are dangerous, anodise black and cover all the optical path you are not actively working on (There may be something to be said for totally enclosed beam paths with O rings at the optics in the manner of a vag burner optical plate).
10: This is at least a grands worth of fiddle on top of the obvious parts (even if you are not counting your time), so maybe a five grand or so worth of light source by the time you build all the prototype sub assemblies.
Regards, Dan.