Bypassing a capacitor helps in a few ways. It helps to dampen the inherent oscillation (energy dissipation) of the initial capacitor by operating closer to the initial capacitor's resonant frequency. Operating closer to its resonant frequency means that it has one a smaller size and is more readily able to dissipate the initial capacitors resonance to the higher resonant frequency of the second capacitor. (Capacitors are lossy) Dissipation is also dependent on circuit topology and conditions
The inherent oscillation is determined by a few things, the physical size of the capacitor, the gap between layers, materials used (paper/plastic type), the capacitor construction (slits/internal shorting at various frequency ranges), electrolyte resonance dampening properties, and many more which I havnt thought of. This is then all put into the circuit which is tuned via various resistance, inductance and capacitances.
Bypassing a capacitor also helps since when In circuit, often the smaller capacitor has to operate on a smaller frequency range and thus has less energy to manipulate (DC). (Think LPFilter vs HPFilter) Fewer signals are arriving at similar moments and interacting on the surface of the capacitor, adding, subtracting as a circuit is full of bottlenecks and tanks.
Bypassing also helps by limiting the vibration of the capacitor layer gaps, thus increasing/decreasing its inherent capacitance. This is because the smaller capacitor is more able to dissipate the frequency of the vibrations at the gap distance.
Of course these comments may be completely incorrect but are the deductions made through reading.
One thing to consider though is that you will not have as seamless a signal as you will now be operating the signal using two or more caps. The recommendation I have read is to use no more than 1/100 of the initial value in uF.
Edit: It is worth it. Considerably
