There are a few limiting factors. At the extreme limit is the speed of light. 186000 miles per second, so at 1ghz you can go .000186 miles, or 1 foot. Electrical signals are slower, so at some point, you just can't get a wire to carry a signal across the whole chip in one clock cycle.
Before that comes the transistors. A "big" transistor takes time and current to switch. As I understand it, you can increase the voltage and that decreases the time required. However, it raises the current (and thus heat), and at some point, the electricity will just spark across the transistor and destroy it. Another thing to consider: transistors havesome cutoff voltage for on/off. in a 5V 386, this might be 2.5V (I have no idea). normally, signals would be .1v for a 0, and 4.9v for a 1 (also a wild guess). if you run it at 50V, leakage currents might mean that a 0 is over 2.5V, but the transistor would see that as a 1. Think of a light switch - you can only turn it on/off so fast. If you go too far, you'll just break the switch.
Also, the clock does not tick at the same time at all parts of the chip. Lets assume that in a 386, the "clock skew" as it is called is 1ns. If you try to run the chip at 1ghz, some parts of the chip are a whole clock cycle ahead/behind others. That is a bad thing
There are a lot more things that come into play, but I think that is enough info for now