There are some requirements for fundingfunding¹², a degreedegree³, tenure and similar for which first authorship and onlyliteral first authorship counts (unfortunately I cannot cite a specific example). If one of the equally contributing authors gains an advantage from being the first author due to this while the other one doesn’t or has a smaller advantage, it can make sense to have the order deviate from the alphabetic one.
Another conceivable scenario would be that the first authors is well-known in the respective field and was made first author to attract a little bit more attention to the paper.
¹ For example several faculties in Germany have schemes for evaluation and publication-based funding that assign special value to first-author publications. Some of those do not mention joined first authorships and do not make sense with multiple first authors, which indicates that *first authorship* is meant literal ([example](http://www.uniklinikum-jena.de/MedWeb_media/Downloads/Fakultaet/LOM+Kriterien.pdf) in German, search for *Erstautor*). ² [This journal](http://www.geoscientific-model-development.net/for_authors/financial_support.html), e.g., lists a handful of funding organisations that will pay the publication costs, if the first author is funded by the respective organisation. Joined first authorship is not mentioned. Even if this may be dealt with on a per-case basis, just flipping the first authors may be easier. ³ For example, for a publication-based PhD thesis, it may required that the included publications be first-author publications ([example](http://www.studservice.uni-kiel.de/sta/5.2.pdf), again in German) without the case of joined first authors being considered. While the latter may be allowed on a per-case basis, flipping the authors may be much easier and avoid a lot of bureaucracy.