The bars are (well, not completely accurate but the easiest way to understand it) spring steel. The same principles as coiled springs are involved, except they're not coiled. The spring rate is a function of the properties of the metal, the thickness and often the length of the bar.
If they are not returning to the OEM position under a standard load (ie: no big bumper, winch, etc) then the metal is fatigued past it's yeild point.
Inverting (well, flipping them back to front or left side to right side) them may help for a short period, but if it is past yield they will eventually "sag" the opposite way.
It is highly unlikely they will "snap" from a simple reversal of loading. A torsion bar break would be caused by a defect in the bar or damage to the bar. For example: you hit a hazard and create a gouge in the bar surface. This causes a stress riser that concentrates stress at that point. It's not designed to deal with a focused point of stress so the metal yields and breaks.
By inverting the bars, you'll probably just speed the rate of sag. Think of it as grabbing a piece of metal and bending it back and forth. Eventually, it gets easier to bend it at the stress point and if you keep at it it will break at where you have been bending it. A torsion bar is going to take more cycles than you can give it before it will break, but you can get it to a fatigue point where the metal properties start to change (ie:sag)
