Heres a simple idea that potentially could be used for replacing transformers. The idea here works only for DC but might be able to be adapted for AC.
First, the idea of isolation:
----S1---+----S3---- | T |
----S2---+----S4----
T is "temporary storage" that acts as the power source while the circuit is disconnected from the load.
When S1 and S2 are both off and S3 and S4 are both on one has "isolation". If these were relays or switches then true isolation would be working(even better than a transformer in many regards).
With mosfets one gets "virtual" isolation that in general would work well.
The main problem is if a switch fails. To reduce the potential for shock one would "sense" the current through all the branches and make sure that current flowing hrough S1/S2 and through S3/S4 are mutually exclusive. There would be a control circuit that would deal with the switching and saftey.
If T is a capacitor another potential problem is the large number of cycles which may reduce it's lifespan. The switching frequency is inverse to the capacitor size. Also one generally would have filtering after the isolation or parallel two isolation circuits out of phase to smooth the output.
Now the great thing about this is that T could also be a SMPS topology(or one which could be modified) so that voltage translation can occur quite easily.
Also for improved isolation one could add more switches and/or put several in series. (while the likelyhood of a switch failing increase it reduces the likelyhood of a catastrophic failure)
It seems though that, as always, someone else already has come up with it!
So, my question is, how realistic is this? I imagine I can modify it to use in my own circuitry as long as I don't reproduce the circuit in the patent?