Glad you got the "case flex" myth sorted out once and for all. Regarding spur gears vs. helical gears: There is some misunderstanding about overlap ratio and how significant this really is to overall gear strength. Not to knock you, because I also used to believe that the higher overlap ratio of a helical gear (2.0+) meant that it was stronger than a spur gear (1.35+), however other more important factors are involved.
If you look at how the helical gear engages, you will see that it starts engagement from one side, and ends on the other side of the individual gear tooth. Even in the most ideal conditions (totally fixed gears), one can visualize the load exerted on one side of the individual tooth, progressing to the other side as gear mesh continues to the next pair of teeth. This motion is how an individual tooth is torn off at the root, despite the fact that two gear pairs (2.0+ overlap ratio) are theoretically engaged. Add to this equation the fact that the idler gear needs to freewheel on a needle bearing whenever another gear is selected, and the fact that required needle bearing clearance = slop = deviation from ideal gear engagement, and you see how the problem is increased.
Now visualize the individual gear tooth in a spur gear pair. As engagement occurs, load is distributed across the entire width of the gear tooth, rather than load starting at one side of the tooth and working toward the other side of the tooth. If you can visualize this, you now see why "beam strength" of the spur gear tooth is greater than that of the helical gear. Beam strength becomes a much more important factor in overall gear strength than overlap ratio in most race applications.
Now consider one more issue: Shaft flex (which you have correctly pinpointed as being FAR more of a factor in Subie trans weakness than case flex). When the mainshaft &/or pinion shaft flex, and the meshing gear pair spreads apart, the ideal design strength of the helical gear pair is further compromised (loss of ideal mesh), while the spur gear's mesh is barely affected. Straight-cut gears that move apart from one another still remain in near-full engagement.
The times when helical gears are "better" than spur gears are:
1) Street applications in which strength is not as much an issue as quiet-running.
2) Road race applications in which gear strength is not borderline (not the case with the puny Subie tranny).
3) Dragrace starts in which some of the shockloading is absorbed by end thrust with a helical 1st gear, vs. the spur gear (in which minimal end thrust is available to help deflect these damaging forces). The question "which is better -- helical or spur 1st gear in drag race applications" can only be answered by practical testing, not theory. And it seems that with the Subie transmission, high quality spur gears do just fine in 1st gear position. So that really leaves us with #1 above as possibly the only time that a helical gear is "better" than the spur gear.
This all pertains only to the main gear teeth themselves, since spur gears can be synchronized (although almost all spur gear manufacturers also have opted to adapt the "dog box" design, eliminating synchros alltogether), and helical gears can be "dogged".
Hope this info helps ---