If you have a temp controller (digital switch) installed you can place the probe where ever you want. If you want the freezer section to freeze you can make it happen. That said, you won't need to.
The whole freezer thing is just a joke from the earlier cited article. He is using a dorm/apt fridge just for ferm'ing, if I recall correctly. You may have to trick/disable the original sensor if you want to crash at maximum, though. I am having that issue right now while trying to crash with a fridge, since the fridge thermo prevents if from running even though it hasn't reached the controller set point. I could cause the evap to ice over if I override the sensor, though. All the more reason to get another chest freezer and mount a fan inside for better transfer.
The electric heater will be more efficient that the cooling cycle of the fridge. Resistive heating is very efficient; energy goes in and is converted to heat. A refrigeration cycle relies of phase changes and mechanical compression ( both larger efficiency losses).
I question that, and submit as evidence the heat pumps commonly used for domestic heating. They are rated as being more efficient than resistive heating, and resistive heating is ~100% efficient. Heat pumps are only more efficient than resistive heat when used in temperate areas, or with ground source heat exchange for colder climes. The Carnot cycle doesn't make cold, it only moves heat. Look up Carnot cycle and COP. It would depend on the efficiency of the motor, though, so a really old fridge may not break even, even if its ideal COP is >1.
Another consideration: if the ambient temp is very near or exactly on your set temp, the resistive heater would be a much better thing to have operating. You could potentially run into a short cycling situation there. If your controller has a cycle timer, that may be a good option to use if that situation comes up.
In this scenario, there isn't really an issue with short cycling, in the sense of coming on very soon after a run cycle, or 'hot' cycling as the more proper term, which is what the Anti-Short cycle Delay (ASD) setting is used to prevent. It will also actually cycle less often, since the heat transfer to ambient will be very slow.
As long as the sensor is in/on the wort (or wherever, but on the vessel wall is best) with a similar diff setting, the length of the cycle will be similar regardless of ambient, although the COP will be more efficient at lower ambient temps, so the compressor won't have to run as long to reach temps. Probably a small difference in real world use, though.
Running for short times isn't really an issue. Running for many periods of short duration, instead of fewer and longer ones, is an issue. That can really only be solved by the increasing the temp differential setting, which would be a trade-off regarding ferm temp stability. Adding more passive thermal mass will also help.
With refrigeration and non-active ferm periods, there is a weird control situation when the ambient is barely higher, or within the setpoint + diff. The wort spends much more time at the upper end of the range before triggering, throwing off back of the envelope calcs to determine avg temps. As long as your diff is small enough, it shouldn't be an issue. If it is, there are medications that can help with that.
That said, you can get much more stable and precise temps by using resistive heat in cold ambient, instead of refrigeration in hot ambient. This is because the heat input power can be varied much easier, and the temp diff can be made infinitely small if an SSR is used. It can even be PID'd for even tighter control, but again, there are meds that can help with that.