Based on this first paragraph:

An edge case is where the program logic meets a boundary condition. A corner case is where you meet more than one boundary condition at once.

I'm tempted to supplement the second paragraph with:

Corner cases may also be ones that seem impossible when they actually are encountered, which in turn consumes a lot more time in the fixing (due to a larger number of false-starts and irrelevant quadruple checks when debugging), and ultimately causes the hardest forehead-slapping once they are finally solved, as in "who would have thought that these two simple, unrelated conditions could have produced such a bizarre result???"