You don't need to put diodes close to the cube, if you use collimated beams ..... just collimate them exactly equal, and the distance from the cube become ininfluent (only remember the polarization, one must be vertical and the other horizontal)
If you try to use the bare diodes without focus them before, the result is worse (some more light lost at the borders, and also you need to keep them practically in contact with the cube, and the same also for the collimation lens)
So, the asembly suggested you from Niko will be ok, as you see, it also have the space for the lens nuts ..... just put the diodes in the right polarization plane, focus them for the same distance, then add and align the cube and the FS mirror
..... i only suggest Niko a modification, to use threaded tubes for keep in place the diodes, instead press-fit them, first cause so is more easy to change them if accidents happens
, and second, so is more easy to rotate them for get the better polarization plane position, before fit them (and a careful made fit like that, is also more efficent in thermal dissipation
).
BTW, Niko ..... You can combine different wavelenght, if you want, if the cube coating is broadband, the only thing to respect is that one must be polarized horizontally, and the other vertically