Unfortunately the darlington outputs used in L293 and L298 devices increase current at higher temperatures - however you are using them as switches, so the current ought to be determined by the load. In this case its the saturation voltage's response to temperature that matters - might be a graph on some of the datasheets, offhand I'm not sure how the saturation voltage behaves (actually its one Vbe drop plus one Vsat - however both are resistance-dominated at high current so its not simple). If the saturation voltage increases with temperature they can be paralleled without worry (current is steered away from hottest devices), if V decreases with T then there could be issues (mounting all devices on a common heat sink can help if the effect isn't too large)
Offhand I'd say you're always going to be better off with a MOSFET H-bridge anyway since the power losses can be much lower at 12V than any darlington output device. For 12V operation you have the possibility of using 2 n-chan and 2 p-chan devices, each driven from a 12V MOSFET driver chip.
But be careful, this sort of design gets hairy quickly as the power levels go up - any mistake can fry (usually explode) a MOSFET or two...
Common mistakes are not driving MOSFETs hard enough to switch fast (necessary for PWM operation) and allowing shoot-through (lower and upper devices on the same arm of the bridge conducting simultaneously - this will definitely cause device failure in a very rapid and expensive manner).