The subject line of your post was "When to greenlight a GDD," which didn't make sense to me. Curiosity as to what the subject line meant is what prompted me to open the thread. The reason the subject line didn't make sense to me is that greenlighting a project is not something a designer does himself. Now that I've read your real question (quoted above), I understand what you're asking. There are three questions: when is a design ready, when to re-evaluate it, and when are its flaws recognized.
The way you used it in your subject line, "greenlighting" seems to be the decision that work can begin on the game. That is a judgment call, and there is no clear "when." I suppose one can deem a GDD ready when the key gameplay is determined, so one can build a demo.
Re-evaluation is something that should happen multiple times. If using Scrum, it should happen after every sprint. It should also happen once a playable demo is built.
Lastly, the realization that it's "too complicated" usually comes after working on building the game for a while and the process is bogging down.
Ah, i see what you mean. I think the simpler explanation would be, at which stage of the 'prototype' could you be turned down by your vision. Since your vision is an abstract idea of a technical formulation i presume.
However, in certain post-mortems you can read the developers stating something along the lines 'i never imagined it work out like this'. Which is quite an interesting statement for the holder of the vision. So where do you draw the line?