In the first version of this article, arXiv:1005.4081v1, we had incorrectly concluded that wave packet decoherence may be observable as averaging of oscillations. This conclusion was based on incorrect estimates for the size σ x of the neutrino wave packets-we had assumed σ x to be comparable to the spatial localization of the neutrino interaction vertices, while in fact taking into account the less tight temporal localization of the production and detection processes increases σ x , and thus also the coherence length L coh , by several orders of magnitude. We are indebted to Evgeny Akhmedov, Georg Raffelt, and Leo Stodolsky for pointing out this mistake to us. Below, we attach the incorrect first version of the article. It will be rewritten to explain the aforementioned problem and address related questions.