Hello Aahz! It sounds like you are talking about the reaction of FICO 8 to the number of cards showing a balance. But our OP isn't interested in the FICO 8 model. He's asking about the reaction of the mortgage models. Thomas Thumb has done some pretty extensive testing and researched it other ways, and he's concluded that the mortgage models are far more sensitive to this particular factor.
In our OP's case, 100% of his cards now show a balance. With SouthJ's proposed approach, he'd be reducing that to 50% of his cards. SJ's approach would also reduce the individual utilization of the remaining card from 50% to 39%. So there would be a gain from the individual scoring factor too.
A fair guess is that the points our OP will get from reducing his cards showing a balance from 100% to 50% is more than the points he'd get from reducing the big card from 39% to 29%.
Of course, nobody knows for sure, it's just in our opinion the most reasonable gamble given the acute timeline as described.