Thanks for the ideas.
I'm quite experienced coding in OpenGL but Slick is new to me, so I often go about to reinvent the wheel and afterwards find out, Slick knows how to do it natively

I guess, the wiki needs to get more detailed about the built-in possibilities, because I often find it quite straining to download and search code examples only to get the grasp of a concept. Maybe I'll contribute some entries / tuts when I'm more experienced, as I'm really convinced of Java and Slick is one fantastic library.
About your answers:
Problems coming down the road is true enough, although rendering to the borders is, in my opinion, not one of them because the whole idea of the concept is not to use getWidth/getHeight in the game calculations at all - those values are fixed, the graphics are done on one rectangle and the output is scaled to be the full window, always.
Losing input seems more severe (you're talking mouse input here, aren't you?), but should be avoided by careful coding.
Until then I use a workaround solution that fixes all problems at the moment:
(oops, that was the submit button, not the code button)
Code:
double xScale = gc.getWidth() / xRes;
double yScale = gc.getWidth() / yres;
[..]
public void render() {
GL11.glScaled( xScale, yScale, 1 );
[..]
}
To handle mouse input I divide the mouse positions by xScale and yScale before interpreting them, thus coming back to game coordinates. This might get a bit inexact when pixel-exact picking high resolution sprites but for my 32x32 sprites it works quite well.
Of course, if the window could change dimensions I'd have to update the scale factors as soon as the window changes, but at the moment it works great.
I didn't try ScalablaGame yet as I spent the day drawing sprites yesterday so I could do more in-depth testing but might in the future, although I can't see the advantages if a simple glScale has the same effect - maybe I'll be enlightened by the effort (but then I'm using a voxelspace approach to avoid futile tests in-game and thus am independent of the screen coordinates all the time I'm not drawing...).
Of course there might be some operations lurking in the API that change the scaling to their liking, but I'd honestly expect a well coded library to push and pop external attributes which need to be altered for a single operation.