Dark horse papers

A journal's impact factor - the number of citations it receives in a year divided by the number of papers published in the preceding years - is often used as a proxy for the importance of the papers it publishes. But the impact factor is a poor predictor for individual articles, since article citation distributions are heavy-tailed, so their mean is strongly affected by rare outliers that receive many more citations than a typical article. 

It is difficult to estimate the impact a paper will have before it is published. Thus, one can find papers published in top journals that after several years have only attracted a handful of citations - the editors and referees overestimated the impact the article would have. Similarly, there are papers that were only published in a specialized (e.g. local society-run) journal and ended up having a big impact. Some examples:

S. Aubry and G. Andre, Analyticity breaking and Anderson localization in incommensurate lattices, Ann. Israel Phys. Soc. 3, 133 (1980). A conference proceedings article with more than a thousand citations and even its own wikipedia page, influential as a simple analytically-solvable toy model of a localization transition.

T. Fukui, Y. Hatsugai, and H. Suzuki, Chern Numbers in Discretized Brillouin Zone: Efficient Method of Computing (Spin) Hall Conductances, J. Phys. Soc. Japan 75, 074716 (2006). This is essential reading for anyone who wants to numerically compute Berry curvatures, since it solves the problem of how to fix a smooth gauge to compute the k-space derivatives of the Bloch functions. Also has more than a thousand citations.

M. Fujita, K. Wakabayashi, K. Nakada, and K. Kusakabe, Peculiar Localized State at Zigzag Graphite Edge, J. Phys. Soc. Jpn. 65, 1920 (1996). This was a paper that was ahead of its time, showing that certain edge configurations of graphene give rise to strongly localized edge states. This was more of a theoretical curiosity, until samples of graphene were isolated a few years later, as you can see in the time series of citing articles:

 

These are just a few examples I've come across in my own research. There are many more out there! Do you have your own favourite example?