that does kinda feel like premature optimisation. First make a game. Second make a game that people want to play. Third make a game that people want to hack. Then maybe look at stopping them from doing that.
Knowing and designing around the unavoidable wave of cheaters is extremely important when building online games.
If you build a multiplayer game then come back to "deal with cheaters" only a few months before launch (or heaven forbid, after launch) you're probably going to end up taking the easy route by using invasion ~~spyware~~ DRM solutions, which don't stop cheaters once it's broken. To contrast, designing the game around cheaters means you might actually implement sane networking design, such as nearly everything being server-authoritative, aka. "never trust the client". If your client is broken in to, the best your cheaters can do is script the game (scripting cheats are currently the only ones available for Dota2 and probably League of Legends) or see a small amount of information that's otherwise unknown (for dota 2 you can see which enemy illusions are fake, for example).
Of course this approach can be taken too far, eg. in Call of Duty: Modern Warfare (the new one) sometimes your camera positioning can lag due to packet loss; this probably stops some aimbots but things like camera orientation probably shouldn't be handled server-side.
And, as with premature optimisation, mantras don't communicate any of the subtleties around the fact that many concerns can influence your code at the same time.
Hard to resist that nagging thought, though.