Try this:
Even if you have a Perspective Camera, attach a small script to your camera, and set the camera's .transparencySortMode to TransparencySortMode.Orthographic. You can probably do this in Awake or Start.
(You can also let some kind of manager do this for you. I actually don't know if this setting serializes into the scene/prefab.)
What this does is make it sort renderables according to flat-plane-distance instead of actual diagonal distance from the camera point.
Let us know if this solution works for you!