If you're going to retire in N years, then that gives you an upper limit to how much you can benefit from learning (or even experimenting with) a new and potentially better approach, versus sticking with what currently works for you. As N gets smaller, the cost-benefit analysis increasingly favors the latter.