This is related to an earlier question (Understanding U.S. Census MSA to place relationships?), on which I'm still unclear. Can MSAs (CBSAs) split counties?
According to Glaeser et al. (Growth in Cities, 1992, p1135), in 1956:
In some cases, an SMA* contains several counties; in others (only in New England), several SMAs split a single county.
*SMAs were MSAs of the time, and Glaeser et al. compare 1956 SMAs with 1987 MSAs.
But as of the latest U.S. census delineation file, I count (perhaps mistakenly) not more than one CBSA (Metro- or Micro-politan Area) per county, even in New England. This conforms with what the U.S. Census says, that "Counties or equivalent entities form the geographic 'building blocks' for metropolitan and micropolitan statistical areas throughout the United States and Puerto Rico." This could be because the U.S. Census's methodology has evolved on that score since 1956. For example, the latest methodology could prevent splitting counties by assigning ambiguous counties entirely to the 'dominant' CBSA (dominant in terms of commuting patterns).
Is it theoretically possible for contiguous CBSAs to split a county (especially an 'outlying county', per U.S. Census terminology)? This would also imply that contiguous CBSAs are possible.