- The decision should still be reversible, so that you don't paint yourself into a corner for future changes in the requirements.
- The simplicity should be weighed carefully against the flexibility and the cost of changing later. If the simplest solution costs 20% less now, but will likely cost 30% more over time, then it may not be the best choice. You have to decide how likely likely is.
I would stress that even with these caveats, one should still be very careful breaking the rule. If you don't have a strong case for a very real risk of ir-reversibility problems or future costs, you should just stick with the simplest solution.