If I understand hinkley's point correctly, it's the constant multiplier (and, in some cases, lower-order terms) that conventional complexity analysis handwaves away as irrelevant because it only looks at the limit as input size goes to infinity.