On light loads the energy required to be transferred is low hence, the duty is low. This of course means that the MOSFET is on for a shorter time period and the energy stored by that inductor during charging is released into the secondary with plenty of time to spare before the cycle repeats. The picture below shows a switching cycle with varying loads. For clarity only the primary and secondary currents are shown and it is assumed to be a 1:1 transformer: -
Picture source - hit the "flyback DCM" button.
As you might imagine on very light loads the charge time is very short and the flyback time is equally shortended. This leaves plenty of time left (hold time) doing nothing however, there is always a little bit of residual energy left in the inductor and now that the MOSFET is off and the secondary diode is not conducting, the only path for this residual energy is to form a tuned circuit with the MOSFET drain-source leakage capacitance (several hundred pF usually).
So the waveform you see is a decaying sinewave formed from the residual current in the transformer (aka residual energy), the magnetization inductance of the primary and the DS capacitance.
It's very normal to see and only happens in discontinuous mode (DCM) because neither the MOSFET nor the diode are conducting in the hold period.
I assumed this type of circuit when explaining the above: -
This is to explain the double charge cycle followed by the extended hold period (pulse skipping): -
- H means "hold"
- C means "charge"
- F means "flyback"