Friction Issue with Multiple Colliders Touching

EDIT:

Sorry about the image links being broken now; they were removed from where I was hosting them.

---

I've encountered an issue with multiple colliders touching, primarily evident when I am making a wall out of stacked boxes. When jumping against this wall, strange things will happen depending on the number of colliders that the jumping figure is in contact with at the start of the jump.

The following images illustrate my test. I have a figure (green) that can move side-to-side and jump, and has no physics material. The red boxes do not have rigidbodies, only box colliders. The text on the floating red box shows how many boxes occupy the same space (later images are tests when several boxes are duplicated in the same space). The text in the upper left shows the recorded top height of each jump.

First test: Image Missing

Second test: Image Missing

Another plain jump, but introducing the blue box collider I've used for this figure so it could jump against walls. This box collider has a frictionless physics material attached (all values on it have been set to 0, and the friction mode is minimum).

As you can see, the maximum height is the same.

Third test: Image Missing

Jumping while touching the single cube with friction. As expected when two colliders with friction run against each other, the jump isn't as high. The height value for this test, and other tests while there is friction, vary slightly.

Fourth test: Image Missing

The results of this test actually surprised me. This already presents an issue with friction in the physics system. The jump is actually higher than a plain jump! The height values for this test, and other frictionless tests, are consistent when repeated.

Fifth test: Image Missing

While I noted in test 3 that the value varies slightly each time it is repeated, it is consistently higher than any height produced in Test 3.

Sixth test: Image Missing

This is where things start getting really interesting. This is similar to the scenario I presented at the beginning, where a wall is made by stacking boxes. The jump height is noticeably higher than a plain jump.

Seventh test: Image Missing

And this is where all logic seems to go out the window (to me at least). If I duplicate the floating box 4 times, so the figure is in contact with 6 boxes simultaneously when it jumps, it goes much higher, and that is with friction! True, it goes up a bit slower and comes down slower, due to the increased friction, but it still isn't reasonable it actually goes that high.

Eight test: Image Missing

As you might have guessed from the pattern in the previous tests, having the frictionless component in contact with the 6 boxes allows an even higher jump. Not shown, I extended the boxes massively to see how high it would go. It went over 9 units high. Also noticeable, even though it is supposedly using a frictionless material, the rise and fall was slower than when it was not in contact with anything (if, at the top of that 9 unit jump, I moved the figure so it no longer contacted the boxes, it fell faster).

I hope I have sufficiently presented the issue. If anyone understands the issue, and better yet, if you know how to solve it, please let me know. Or if you think this should be submitted to a bug report or other location, just tell me where to go!