The ability of a transistor to turn on and off (which is really what we're looking at) is heavily governed by the capacitance which is caused by a voltage on the gate of that transistor (which can be thought of as the button on the switch... if people have taken circuit design before then please, I apologize for the gross over simplifications). This capacitance is also governed by a number of things, including the mobility of the electrons in the Si compound and the mobility of the SiO2 (insulation) layer. This is sort of the ability for electrons to flow through the wafer without incurring resistive forces (due to scattering and electrons bumping into things)… Anyway, long story short, this mobility has to do with temperature and once the temperature goes up a number of funny things start to happen:
Channel length modulation in the pinch off region of the transistor (directly underneath the gate) gets more profound and as such this causes the transistor to deviate away from its ideal characteristics.
Also, the parasitic capacitances (namely from source to drain and from gate to source) become more pronounced and so that when a transistor wants to raise or lower the voltage from gate to source and from the source to the drain it has to “wait” for the voltages on these parasitic capacitances to “drain” or charge. This then may cause the wrong bit to be passed through a transistor during a certain phase of a clock cycle because the transistor wasn't quick enough to pass on its “information” to the next section before the time allocated. This is bad.
So, what we do as overclockers is put more voltage in (the transistor wont necessarily change voltage levels quicker, though this is not always the case, but the voltage thresholds between transistors will become more pronounced as a result) and we keep the temperature down to help with the mobility… ideally we would like to make a super conductor with REALLY low temps. Anyway, mobility of the material and the condition of the SiO2 layers is NOT related to the current flowing through the transistor (or at least the most profound effects dealing with overclocking are not), they are related to temperature. For instance…
Resistors have a wattage rating (and for now we can think of this as the channel in a CPU MOSFET transistor), which is to do with the amount of energy that transistor can safely dissipate. Transistors are actually black body objects (maybe you’ve taken some physics and have studied Plank’s “black body radiation”… this is the same thing) and as such they exhibit the nature of uniformly distributing (over all frequencies… though this isn’t quite true) non deterministic energy dissipation so that if you put current into them (and therefore push in a few electrons with kinetic energy) and out comes heat energy. This is why transistors heat up in the first place. What becomes a problem is when the resistor cannot dissipate enough heat because you put too much current (which is actually more common in the wires of ICs than in the transistors) and the cross sectional wire area actually gets SMALLER! Though this is not intuitive, you then get thermal runaway and the wire space eventually breaks (under the strain of this coming together of wire materials). However, I think that this happens as a function of temperature and NOT current, though since most devices have a current rating people tend to get confused.
The reason why I mention all this is that as long as the temperature is under control and a system is running stable, I don’t think there will be any MAJOR long term effects caused by increasing the clock speed or core voltage on a CPU. CPUs are solid devices. The ALUs, controllers and DRAM, etc,… inside them are hard wired and unlike software they do not change.
