Yes, I made the 802-11-wireless.band setting (bg).
In one test just now, I could not reproduce the problem I had before, but I can’t conclusively prove that that is because of the band setting.
Another way to attack this issue might be to implement 802.11r in the network, perhaps with 802.11k and perhaps with 802.11v, to encourage client devices to switch to stronger signals more readily. (Don’t ask me for details; I know not much more about these standards than you can learn from Wikipedia and your favourite search engine.)
I feel like I’ve derailed the thread a bit here. Sorry about that.