Clock drift is the slow divergence between two independent sample clocks that are nominally the same rate but never exactly equal — a capture device running at 48,000.2 Hz against a playback device at 47,999.8 Hz, say. The tiny per-sample error accumulates: over a long call or stream, audio and video gradually slide out of sync, or an audio buffer slowly fills or empties until it overflows into a glitch. It is a defining problem of long-running, separately-clocked systems like WebRTC and IP audio. The fixes are continuous resampling to gently match the rates, or, more crudely, dropping or inserting an occasional frame to claw the timing back.
