If I am not mistaken, you don't call PShape's scale, but Processing's one. But well, if you specify the dimensions in the shape() call, you get similar results, because Processing internally just calls translate & scale before drawing the shape. Hence the odd result you get.
Perhaps you can walk the SVG hierarchy and alter the internal dimensions according to your scale, but it will be hard...

Thank you for your reply! I just tried calling PShape's scale method and got the same results. According to the reference, shape.scale() applies directly to the shape, so I don't understand why the stroke would morph even after the next draw cycle.... I was really hoping this would be an easy fix :(

From what I see in the code, PShape.scale() and similar just update an internal matrix variable, which is applied to the PGraphics before drawing, so no improvement here, just the same old scale effect.

Can you elaborate on what you mean by "just update an internal matrix variable"?
So you're creating a graphics buffer for scaling?

I would like to use pshape scaling within draw to match zooming in and out, though I can't find how this is done dynamically, interactively.

" So you're creating a graphics buffer for scaling?"
No, the said matrix is a classical 4x4 matrix (or just 3x2 for 2D) used for transformations: scaling, rotation and translation. The coordinates are multiplied by this matrix when drawing. I suppose in JAVA2D mode, it is delegated to the Java graphics system which scales everything uniformly. I just checked, actually, in P2D mode (on Processing 1.5.1), the stroke isn't scaled as in Java2D mode, without much surprise. Alas, the rendering isn't very good either...

I don't understand your second paragraph...

You should read a bit on vector drawing transformations. Again, scaling, translating, rotating and other transformations, combined together, can be represented by a matrix.
When you scale a shape, you multiply the coordinates of the points (vertices) of the shape. When you translate it, you add constants to these coordinates. Etc.
Imagine a simple circle. The coordinates of this shape can be, for example, on North, South, East and West of this shape. Scaling it means to move these coordinates, to enlarge (or reduce) it.

http://forum.processing.org/topic/vectors
http://processing.org/learning/pvector/
http://processing.org/reference/PVector.html

Thanks for getting back to me.  I could possibly use a matrix to buffer the scaling but do I really need to?

I meant to say location, as opposed to coordinates.

For this question there are 3 types of coordinates.
1. the coordinate system of the underlying map the .svg is placed on.
2. the coordinates of the .svg (the location of vectors that define the shape in relation to each other)
3. the location coordinates of the .svg on the map (ie (40.00, -74.00))

I understand that shapes created with specific vector coordinates are changed when scaled.
I do not want to scale global coordinates, I simply want to scale the shape (by using some associative property as a multiplier, like the map's scale rate- such svgshape.scale(.9), to help scale it).

I have an .svg in a file, it already has a certain scale.  I place the shape on a specific location coordinate on a map. The location on the map doesn't change. 
However, this is where I get it wrong and the .svg is not only changing scale but also location of the overall shape (it shifts left eventually ending up off screen, when I zoom out, as opposed to staying center) .

Does the object.scale() function scale from the corner, not the center?
This must be relatively simple, no?

Cheers,

I mentioned the matrix because of the original topic. But you don't need that to make transformations, of course.

As for your specific problem, I suggest to make a simple sketch illustrating your issue, perhaps you make a simple error easy to fix.