Same problem as Q1 but with larger constraints (up to 10^5). The O(n) approach with suffix min and prefix max is required. Precompute suffix minimums. Scan left to right with a running prefix maximum. Return the first index where prefix_max - suffix_min int: n = len(nums) if n == 0: return -1 suff_min = [0] * n suff_min[-1] = nums[-1] for i in range(n - 2, -1, -1): suff_min[i] = min(suff_min[i + 1], nums[i]) current_max = -float("inf") for i in range(n): current_max = max(current_max, nums[i]) if current_max - suff_min[i] <= k: return i return -1 Step through this solution on TraceLit — set breakpoints to see exactly when the instability score drops below k.