When the clock edge arrives, a d flip flop doesn't update Q instantly. The clock-to-Q delay tcq is the longest time after the edge before Q is guaranteed to show the newly captured value. It plays the same role for a flip-flop that the propagation delay plays for a gate.
Its partner is the minimum clock to q delay tcq(min), the earliest Q could start to change. Between tcq(min) and tcq, Q is uncertain.
Why it matters: in every register-to-register path, data can't even start through the logic until it has left the launching flip-flop. So tcq is the first term of the setup constraint:
T ≥ tcq + tpd + tsu
Forgetting it is a common slip that makes a design look faster than it is. Some textbooks write tpcq for this number.
Note how tcq differs from the two input rules, setup time and hold time. Those say when D must be steady; tcq says when Q becomes valid. All four numbers are measured from the active clock edge.