First, imagine a dc voltage driven into the trace after some rising edge. The edge travels down the trace and gets reflected and travels back to the source.
in the time before the edge makes it back to the source, the load on the source is the characteristic impedance of the trace (eg: 50R). But after the wavefront reaches the source, the load on the source is equivalent to the load at the other end of the trace (eg: open circuit or a high impedance termination). Before the reflection returns to the source, the driven signal is distorted by the source loading of the characteristic impedance.
This travelling time is obviously dependent on the trace length. If the traveling time is a significant fraction of the total pulse width time, then the signal will be distorted for a significant fraction of the period. If the time of flight of the signal is small compared to the signal period, then the distortion will be only for a short time after the edges.
It gets worse if the signal is reflected so that the reflections are logic high when the source is trying to drive logic low and vice versa. Then the source is essentially driving into an output in antiphase all the time. (This is how damage is done to transmitters when not properly terminated.)
Now, make the slight mental adjustment to sinusoidal waveforms. If the trace length is such that the reflections are 180 degrees out of phase, then the source is driving into the worst possible load (short circuit). if the trace length is doubled, then the reflections are in phase and the source sees no load.(This is ideal, and can be an alternative to termination). if you increase the length a small amount, the reflection will be almost in phase, and the source loading will be minimal, so it's very close to the ideal no loading situation. But of course, a full wave length plus a small section is the same as just the small section, so a very short track is close to the ideal no loading situation. How close is dependent on the ratio between the trace length and the wavelength. At this point, all good engineers just use a rule of thumb rather than analysing each case. Mine is 1/10th of the wavelength., if you have rise time instead of frequency you can use f = 0.35/tr