Proof by induction:
Start off with a single rectangle A with one integral side X and one non-integral side Y. Add another rectangle B to it, also with one integral side and one non-integral side, to form a third rectangle A + B. Trivially, B must have the same side length as A on the side upon which it is placed if A + B is to be a rectangle.
If B is placed onto side X (i.e. one of B's sides must be of X's length), then A + B must have one side of length X - i.e. one integral side. So A + B has an integral side.
If, on the other hand, B is placed onto side Y, then one of B's sides must be of length Y - i.e. non-integral. Therefore, B's integral side must be directly aligned with X. So one of A + B's sides is given by the length of B's integral side plus the length of A's integral side, and so must be integral itself. So A + B has an integral side.
We see now that in either case, adding a rectangle with one integral side to another rectangle with one integral side in such a manner as to produce a third rectangle will always produce a third rectangle with one integral side. This process can be repeated ad lib to produce any type of composite rectangle (i.e. adding C onto A + B, then D onto A + B + C, and so on to arbitrary complexity). Therefore, since any composite rectangle of the type described can be produced by this method, and every rectangle produced by this method has one integral side, every composite rectangle of the type described has one integral side.
I can see why it's one of your favourites - quite a nice solution! Sorry the proof's so long - mathsy language is the easiest type to use when working this sort of thing out, and it's too late for me to translate to English...
____________________________
Stupidity kills.
Absolute stupidity
Kills absolutely.