I think personal verification is a rather unrealistic standard. If we used that metric for acceptance of answers, we'd have a much lower acceptance rate. e.g. this question -- I have no realistic way of verifying either Aaronut's or Darin's answer. I don't know how to test for ethylene gas, nor do I know how to verify Darin's claims, which are backed up by McGee. At some point you really do have to trust a source. If users want to use this as a personal metric fine, but I don't think it's advisable to present this as any kind of "suggested" behavior.
I also think it is a little melodramatic to infer that we would look like fools if an accepted answer is wrong. Acceptance, as well as votes, can be changed quite simply. If an answer is given, and provides reasonably trustworthy references, there is no harm in accepting it -- even if it later turns out to be wrong. If someone who knows better comes along and points out the error, that's great; we can thank them and make corrections.
In the banana question case -- I'd speculate that the vast majority of expert chefs don't even know the answer to that offhand. If inclined, they'd likely do similar research to my own, and accept the explanation given by the reasonably trustworthy USDA and Canadian equivalent. I've seen chefs I deeply respect say things that are completely wrong on television, and in print, and I don't think them fools. They just can't be expected to have the necessary chemistry/physics background to answer some questions truly accurately.
For example, caramelization and the Maillard reaction. There is a lot of misinformation, confusion, and misquoted foo out there, even on this site. Repeating what you've heard or read from a reasonably trustworthy source does not make you a fool. It makes your sources wrong, and that's patently different.
In short, the only grounds to accept an answer are really your personal preference. If you want to accept the most upvoted, do so. If you want to accept a completely wrong answer, do so. As always, I'm firmly against any attempt at trying to regulate, or even encourage the way people use their votes and acceptance.