The distinction interviewers probe: amortised versus worst case. Appending to a dynamic array is amortised O(1) but occasionally O(n) when it reallocates. Latency-sensitive trading systems often prefer a predictable O(log n) to an amortised O(1) with a bad tail - raising that unprompted is a strong signal.
Space complexity includes the recursion stack, which candidates routinely forget.
Constants matter more than people admit. At realistic sizes an O(n log n) algorithm with good cache behaviour routinely beats an O(n) one that chases pointers.