(Not to pollute rolo's 2045 thread further.)
It occurs that there are many ways to do this, some perhaps better than others.
Mirror diagonals are just like newtonian secondaries - so if you open up the mirror, you can place them against a flat to check for accuracy.
Prism diagonals also should have three flat surfaces as well in like kind.
After that, you're left with secondary lateral chromaticism if the glass isn't uniform in refractive index, or perhaps if the angle of a face is wrong.
If the glass is "swirled" or strained, it's non-uniform density will deflect parallel light, much like with air density in a Schlerien system - like when you do a Foucault test with your warm hand causing ripples.
Perhaps DPAC with the diagonal in front of the objective, with the mirror to the side "leg" of the diagonal would get that best. Perhaps use the objective dustcap with a central hole to support the diagonal with?