The cell borders example is misleading. The readability issue is not the cell borders themselves, the issue is that the borders are heavier weight than the text, and there’s no difference between the borders separating the row and column headers, and the borders separating the data rows and columns.
If your only choices for gridlines are “off” and “obnoxious”, “off” seems like a good choice. And for small tables, no borders works well. But for larger tables, finer lines (maybe in a lighter color or shade) can really improve the readability.
Ironic that “Maybe” seems to have one of the narrower ranges of probabilities...