I'm in between on this: design for the general, yes, but at the same time keep an eye out for the more obvious what-ifs and sort them out at the design level where possible. This principle holds true both for game-rules design and adventure writing.It's never worth it. You should always design for the general and worry about exceptions later (or not at all).
This "gentlemen's agreement" is exactly the sort of metagame thinking that drives me nuts.As posted earlier in the thread, if there's a rule that breask the game then the best reaction is to have the mature discussion, 'This rule breaks things, let's all agree to not use it."
On the odd occasion where mature discussion doesn't work then congrats you've just identified one or more obnoxious players. The next step is to consider whether or not you want to continue this gaming with them (maybe change to a different game, maybe kick them, maybe leave, lots of options here).
For example, I am not using forcecage as a box again in 5E. Why? Because it just plain sucks for a player to spend the next 10 turns doing nothing. We've all agreed that my players won't do it to a single foe either, because it sucks for me to spend the next 10 turns doing nothing. If there are multiple foes then they will happily use forcecage (or banishment or similar) to take out one foe while fighting the rest.
Either rules-fix the spell so it works like you and your table really want it to or accept the fact that both you and your players can (and IMO should, in-character) use it to whatever best effects present themselves at the time.
The mature discussion should start with "This rule breaks things, how can we fix it so it no longer breaks things?", and go from there. (personally, were I the DM in this example I'd start one step further back and ask "Does sitting out for ten turns break the game for you?", and if I got any "yes" answers some big red flags would go up; as IMO and IME not being involved for a while now and then is a hard-wired fact of RPG play)
And if it turns out the rule can't be fixed then maybe the nuclear option is to - in this example - drop the spell completely.