I can't say anything Airbus-specific, but the designers of closed loop controllers (i.e., the parts of the AP brain that control the pitching to seek/hold a speed in speed mode, or an altitude in altitude mode, etc.) always have a balancing act to play: If they set it to be too sensitive, it'll correct with too much intensity and overshoot the target value, and even worse overshoot the correction to the overshoot, etc. and have oscillations. If they set it to not be sensitive enough, then it'll be too slow to act and not be able to stop a developing deviation in due time.
Finding a sweet spot of sensitivity between these 2 extremes can be elusive. Especially since the airplane has different behavior in different regimes. So it can be very difficult to make the AP interface with that behavior throughout the entire range. It'll be OK in one range, but be too slow or fast in another range. For example, the CRJ in speed mode in a climb, is very good at not chasing the speed in normal turbulence. It holds a nice steady attitude. But at high altitudes (I mean mid-20's, "high altitudes" right?
) you'll get these long-period, and very slight in intensity, up and downdrafts... even on a smooth day. But the rate at which they come, happens to match the rate at which the AP pitches for speed, in a way that results in the AP chasing the speed back and forth, slowly bouncing in vertical speed from +2000fpm to level flight. It's a resonance, like an extremely slower version of the opera singer shattering the crystal glass.