Is it better to make tests with a lot of portalable surfaces, and only putting nonportalable surfaces where necessary to remove various, unintended methods of completing the test, or to only leave the portalable surfaces necessary to complete the test.

It seems to me that the latter method makes for slightly easier tests, but the first method makes the player feel more like they cheated the mapper and found an unintended solution (with that many possible surfaces, they might have). Also, the second method is much easier to create as you don't need to worry as much about exploits, but it makes the solution slightly easier to find (there are only so many possibilities with only 3 surfaces to worry about).

I know there are other factors that make a test difficult, but this seems to be the easiest one to control.

1) It's up to you.
2) Well if you want to go theme wise. Then broad surfaces in destroyed, and clean, portable surfaces on the necessary in the underground chambers. Both broad and specific places in Wheatley.

It's really a balancing act. Too few portalable surfaces, and the solution is obvious, but too many and the puzzle becomes tedious. I usually leave unintended solutions in if they're not too obvious - in fact, I may put some in on purpose. (Although technically that makes them intended)

Besides, the player feeling they have cheated the mapper can be good in moderation. Too often though and it just makes them feel that the mapper is an idiot. Really, its up to you.