I think everybody who has thought much about it has similar concerns. I'd propose we start soon, start small, ramp up slowly, and be thorough about the data collection.

As uncomfortable as it is to experiment on the only planet you have, even worse would be to wait too long and then, in a panic, try to do everything that might possibly work all at once and to as extreme a degree as possible.