Here's four A4983s under a 40mm chipset cooler. Since my first experiment with heatsinking worked so well, and at the same time showed where things could be done better, I cogitated for a while and came up with the above. As you may notice, they're not nearly finished yet- only two have sprouted enough wires to control, and they're only temporary until I obtain some more headers and some more screw terminals.
The current adjustment pots are sticking out the sides, and I cut channels in the heatsink base so as to not foul the onboard Vmotor capacitors. If ever I remove the heatsink I'll grab a photo, but if you imagine a keyway in the face of the heatsink you already know what it looks like. There's also a 7805 and a large capacitor, and a liberal sprinkling of 100nF decoupling capacitors. You can never have too much decoupling when high frequency, high current PWM and logic signals are in the same vicinity. The motor wires are twisted even over these short runs for exactly the same reason.
Although I haven't measured the new maximum currents I don't think I need forced air, but I prefer having it set up to remove rather than trying to shoehorn it in later.
I think that this may become my reprap motherboard- what more do I need? A temperature sensor interface (max6675 and thermocouple waiting to be hooked up and tested), a mosfet (I have some of the amazing IRL3803s on the way) for the heater and some end-stops (already arrived, ironically lacking 390R). Should be plenty of room for all that plus my arduino on the unused board space, and I can run the lot off the 7805.