Jorgensen’s first law: whatever your planning cadence is, work from one window will spill over into one third of the next window.
For example, if you plan on a quarterly cadence, the loose ends of the last quarter’s work will take roughly a month of the current quarter to tie up. If you plan in months, the bleed will be just over a week. If you plan in weeks, roughly a couple days (the math breaks down a bit here with the weekends).
One consequence of this law is that you cannot actually change direction at the speed of your planning cadence, but at ~1.3 times your planning cadence, which becomes a real issue for quarterly or longer windows.
Another consequence is that, the longer the planning cadence, the more likely the work you plan for the next window is to be impacted and to require re-planning—and the rise in likelihood is worse than linear. That is, if you plan on a quarterly cadence, you are “finalizing” planning for window N+1 while there is still a full month of work to do for window N, any piece of which work could invalidate some assumption you used in your planning. With a weekly cadence, there is only two days' worth of work that could screw you, and the “fog of war” looking out over two days of development is much less than 1/15th of the fog that settles in over 30 days.
A fair conclusion is: keep your planning cycles as short as practical.