"Somewhere in the middle" is one of three possibilities if you reduce everything to a one-dimensional line and pick two points. Even in the rare cases where an oversimplified single continuum expresses the best solutions to address a problem, there is little reason to assume that two point alternatives aren't together in being the wrong direction away from the optimum.

However, I really think golemotron was just referring to when a trend becomes a default that no longer needs to have its case made in new situations. Like assuming middlegrounds are implicitly correct, or cargo culting, it's just a form of assuming there are shortcuts that make understanding the mechanisms of success or doing actual analysis are unnecessary.