Consensus is an imperfect metric. Some of those things, even if discarded by consensus in the near-to-long term could prove valid in a couple years' time.
We want as many avenues of exploration as possible, even once the proof is decided right or wrong.
Moderators on meta.mathoverflow.com decided not to allow possibly premature discussions on mathoverflow: http://meta.mathoverflow.net/discussion/590/. The arguments are quite reasonable: