Assuming that existing applications didn't change the value in the render field - because it didn't have any effect - using the preview value may introduce a tiny performance penalty for existing applications (I take this for granted but I don't really know if that is actually true).
Even though I think that the correct solution would be to use the preview value, because that is what the UI is telling to the user, I'd rather use the default value of 1.0 when the render field is set to 0, to avoid existing programs starting to behave differently.

I'll look into the the versioning thing.

I think this only makes sense if/when the BGE font code get to use the same drawing code that Blender does for the Viewport. There is no BGE-specific doversioning, so this will be complicated to handle (and a high resolution will cost a lot for unaware games).

Ok, this is what I understood so far.
The version file handles transitions from version to version. If an old field that has a wrong/non existing value is found, it is "fixed".
I might be wrong but I'm not proposing to change anything in the blender file, the bge font object is created on the fly by the engine.
If I understand this thing correctly, no transformation is needed from version to version.
I'd also like to point that the current default value (1.0) is the same value applied by the patch.
If for some reason a old blender file - used for a game - has a render resolution value other than zero, then and only then, the program will behave differently.
I don't see how the way the text is rendered - via custom routines or sharing ui code - can make a difference. Right now for each Text object one has to create a binding to a python function that fixes the font resolution, otherwise the text is an unfathomable mud of pixels.
So, to recap, I think:

1. it is not advisable to use the higher preview resolution because it will suddenly introduce a performance hit (users already implemented fixes for the resolution issues)
2. there is no need to add transformations for older files

therefore the proposed patch, in its original form, is (again, imo) correct and useful. Actually, I'd like to change it in the more explicit form:

/* Sets the real time initial resolution for the rendered font */
const float renderFontResolutionU = (float) text->resolu_ren;
const float minimumFontResolutionU = 1.0f;
if(renderFontResolutionU > minimumFontResolutionU) {
    m_resolution = renderFontResolutionU;
} else {
    m_resolution = minimumFontResolutionU;
}

Of course, feel free to use it or not, I'm not trying to shove it down anybody's gullet :D.