I love that they published a paper on this. For most of us, this is one of those things that surprises you once and then convinces you not to use Excel for genomic work in the future, since no magic should be applied without explicit instructions from the scientist. For them, they realized that this Excel/gene name error affected almost everyone at least once, probably reasoning that journal referees would be quite sympathetic despite the otherwise bland nature of the story.

The type of auto-conversion that's going in, as mentioned in the article, is e.g., DEC1 (text) to 1-DEC (date), etc.