This is probably just faulty logic on my part, but has anyone tried this simple experiment? (sorry I have virtually no test equipment at the moment to check this myself)
A capacitor C1 discharges into a series of three capacitors, Co, Cm and Co:


The outer caps Co are large, but the middle one Cm is tiny, maybe just a spark-gap.

Now when Cm is shorted by the closed switch, the total series capacitance C_total is (1/Co + 1/Co)^-1= Co/2. This is still a large capacitance, so C1 can discharge into the series, inducing charges on the two outer Co capacitors.
But when the switch is opened, the middle Cm becomes part of the series, and since it is so small, it makes the total series capacitance  C_total almost zero. Therefore, it loses its ability to hold charge, and so the charge should migrate back to the source capacitor C1.
I can't see why this process wouldn't be repeatable, if the switch opens and closes repeatedly, would it cause a current to oscillate in the loop?

My 2 cents: At a first glance, I have a hard time understanding the polarities the way you've described it, I would like to say that the polarity of C1, as shown in your diagram, should be reversed in order to "complete the circuit", with all voltages adding up to zero as the capacitors continously discharges, as opposed to VC1 = VCo x 2 + VCm.

On the other hand, not having actually tried building such a circuit and measuring the voltages myself, you could just be correct.

Thanks for your reply; my apologies, the colors and polarities weren't significant. It was simply a case of cap C1 has a charge, and it is discharged onto a series capacitor: C_total = [1/C_o + 1/C_m + 1/C_o]^-1 which could be varied by switching C_m  in and out.
When the middle cap is shorted, the C_total is a maximum, so charge can be transferred from C1.
But when Cm is in, and C_total =~0 Farads, what happens to the charges and voltages? You would end up with a zero C_total but with C_o containing a substantial charge, the voltage V=Q/C_total  ought to skyrocket?

If any sort of repeatable charge flow is generated just by virtue of opening and closing a switch, which costs little in energy terms, is that generating energy?