We tested a lot of commercial 90-deg folding flats at work in the as-used orientation, double-pass. So-configured, it is allowable to analytically remove both power and coma, because focusing and aligning/collimating on a star drives those out anyway. The more costly ones were generally better, but still a lot of variations, and an occasional clinker. I recall that the Edmund Optics ones were surprisingly good, for the bargain price. We found about 80% were fine for use. I don't know it they farm them out or build in-house... but it was clear that they must do conscientious pre-ship interferometry before releasing them to customers.
The sizes we had in the bins ranged from piles of dinky ones (2mm-12mm), lots of mid-sized ones (to 5in), handful of biggish ones (to 12in) and a very few very big ones (to ~2ft wide by 3ft long!) The huge ones we made in-house... custom for our own use or contracted customers.
I also tested a bunch of commercial Star Diagonals, with widely varying results. Again, the premium brands (e.g. Astrophysics, Televue, Denkmeier, Lumicon) were wonderful.
The weak link is often not the glass itself, but how it is mounted in the cell/spider. Sometimes just re-mounting with less stress on it can improve it a lot! Also, where they are in the imaging-chain. Close to focus (Star Diag) is more forgiving, because each field point sees only a small footprint on the fold. Farther (Newt Fold) is more demanding, because each field point sees most of the fold. Tom
Example test results on one. [I designed and built the test sets myself.] >>> Tom