MakePixelPerfect on SlicedSprites

[QuantumMechanic]

1.         override public void MakePixelPerfect ()
2.         {
3.                 Vector3 pos = cachedTransform.localPosition;
4.                 pos.x = Mathf.RoundToInt(pos.x);
5.                 pos.y = Mathf.RoundToInt(pos.y);
6.                 pos.z = Mathf.RoundToInt(pos.z);
7.                 cachedTransform.localPosition = pos;
8.  
9.                 Vector3 scale = cachedTransform.localScale;
10.                 scale.x = Mathf.RoundToInt(scale.x * 0.5f) << 1;
11.                 scale.y = Mathf.RoundToInt(scale.y * 0.5f) << 1;
12.                 scale.z = 1f;
13.                 cachedTransform.localScale = scale;
14.         }
15.  

I don't understand the intention behind the way MakePixelPerfect calculates the scale of a sliced sprite.  It appears to be trying to ensure that the sliced sprite has an even scale.  Our game dynamically scales a sliced sprite from a scale of 1,10,1 to a scale of 30,10,1.  To make sure that the sprite looks clean, we calculate the scale and then apply MakePixelPerfect.  This gives a somewhat unexpected result.

Looking at the x component only, since the other components stay constant.

For an x scale of 1:
the x value is halfed making it 0.5.
then Mathf.RoundToInt rounds it to 0 (values exactly halfway between two integers are rounded to the nearest even integer). 
Then the 0 is shifted left by 1. 
Resulting in 0

This generates a LogError -- Scale of 0 is invalid! Zero cannot be divided by, which causes problems. Use a small value instead, such as 0.01

for a x scale of 3:
the x value is halfed making it 1.5.
then Mathf.RoundToInt rounds it to 2 (values exactly halfway between two integers are rounded to the nearest even integer). 
Then the 2 is shifted left by 1. 
Resulting in 4

This creates a sprite that is one pixel wider than the original scale requested.

Its possible I am not understanding an underlying assumption for sliced sprites.  Can you clarify whether this is intended behavior?  My initial expectation is that it would have done something like:

scale.x = Mathf.Floor(scale.x + 0.5f);  // round scale to the nearest hole number, if scale is exactly half -- round upward.