There is also one additional choice that I use on occasion: widen the spread of the tested side such that you receive a credit for the roll and an improvement of the tested short option.
If you've played with options setups at all, you'll know that the wider the strikes between the short and long options of a credit spread, the greater the credit you receive for the spread. Widening the spread slightly on a roll may allow you to roll up the short option a strike or two, such that your probability of profit is increased, but there are also drawbacks to doing this, one of which is that the buying power devoted to the spread increases and therefore the setup risk. Moreover, if one "widening" spread roll doesn't do the trick and price continues to move against the position, you're faced with rolling again and again with potentially widening spreads, greater devotion of buying power, and therefore greater risk to the setup.
Because of the drawbacks of doing this, I only use it when it appears to make sense such as when I think my underlying assumption in the setup remains largely correct but price has taken a temporary bounce/dip against my position, and I do it -- as with all things -- small, widening the spread just enough to (a) get a credit for the roll; and (b) to modestly improve my strike prices.
Naturally, I will also want to sell an oppositional side against the rolled widened spread; I generally do this at my "regular" spread width and not at the width of the widened spread, but you can naturally consider doing that, too, although my take on that is to confine the risk to one side of the setup so that you don't get caught in some kind of whipsaw that proceeds to test the oppositional side you've set up (hey, I've had it happen).
Again, the goal is not to hit a home run with the tested setup; it's to get it back to scratch.