You have a load $Q_1=U_1 C_1$ on the first capacitor, and a load $Q_2=U_2 C_2$ on the second capacitor. The combined capacitors have a capacity $C_1+C_2$, so you'll end up with a voltage of $V=\frac{U_1C_1+U_2C_2}{C_1C_2}$. So far, so nice, but the total stored energy is different. So at some point in that summary view of charges and capacity, we did a bit too much of handwaving.
Now if we connect those capacitors and consider the charges/voltages to immediately adjust, infinite current has to flow. "Infinite" is not an option even when dealing with very good components, so we have to take into account that the connection between the capacitors will have a bit of resistance and a bit of inductivity and is immersed in the ether (well, there is no such thing but empty space is good enough for EM waves). As a result, we have a dampened oscillator which will come to rest (with the voltage we have calculated) when the missing energy has been dissipated. Assuming the connection is solid enough, most of it will be very quickly radiated as RF rather than heat and if you have a radio running nearby, it will likely produce a click at that moment.