In many computational studies, mechanisms appear to have so-called ambimodal transition states, i.e. a transition state which can lead to multiple products, in conflict to the common undergraduate idea of a reaction pathway.
To give an example, Houk, in Mechanisms and Origins of Periselectivity of the Ambimodal [6 + 4] Cycloadditions of Tropone to Dimethylfulvene, presents the following scheme:
Reaction pathway for [6 + 4] cycloadditions of Tropone to Dimethylfulvene. Taken from ref 
In the scheme, TS-1 may either give 7 (the 'expected' product) or go directly to 6, completely skipping TS-Cope-1 (which would allow 7 and 6 to interconvert) — Houk describes this as an ambimodal transition state, that is one transition state that can lead to two possible products.
Although what he says makes sense, it seems to me that, given 6 is the lower energy product out of the two, it would be favoured thermodynamically anyway.
Upon further reading, I discovered the concept of valley ridge inflection, which appears to describe the phenomenon observed above, in which we completely 'miss' the intermediary transition state:
Why do these ambimodal transition states occur, and how does the situation differ from normal transition state theory?
Notes & References
:Houk, K.N., J. Am. Chem. Soc. 2017, 139, 8251. DOI: 10.1021/jacs.7b02966