I have, with little success. However for this particular problem I am using OpenWRT in LUCI, so I do not think this forum is an inappropriate place to ask.
Well, except that GLInet supports LUCI fully. I understand if you are not willing to help me, but perhaps let's leave our discussion there in the hope that someone else on this forum might be willing to help?
As indicated the best (and only) place for your questions is the vendor's own support channels. These forums are for vanilla OpenWrt support, not for whatever mutilated and outdated SDKs vendors use.
There's nothing we can do for you here. Talk to your vendor, they have a forum of their own and other support channels too.
That's a beautiful but crippled analogy. VW and Skoda are almost rebadges of the same hardware. What you are running is an ancient Qualcomm SDK, with a crapload of their own stuff mixed in, and Gl.iNet has thrown LuCI on top of that. So compare it with taking your Skoda Superb to an authorised dealer, with a modern body (LuCI), but with a third party engine loosely based on their own (the SDK) in a configuration the manufacturer (OpenWrt) never delivered it. And the authorised dealer says: 'I'm not familiar with the stuff being pulled there'. That's closer to what you're experiencing.
I'm glad you found a pointer, but that thread you linked to seems to be from someone in the same SDK boat (running 15.05 in 2018...). So yes, it's a Gl.iNet problem, since by now OpenWrt has moved on and this has been fixed for four years.
Anyway - you can replace the hostapd script, it's just a shell script. Or patch yours. But make sure you're not removing other stuff in the process if you replace it. No need to recompile.
But only because I commented out these three lines (688-690) in the heavily Qualcomm edited /lib/wifi/hostapd.sh as the generated rokh had formatting errors or incomplete sections. An what is more, with option ft_psk_generate_local '1' it is possible rokh is not needed
I had a few issues where wifi restarts would wipe parts of the config (especially the nasid) uci commit wireless seemed to improve things some times - although not all of the time.
It seems devices can now connect but I need to find a way to test if fast transition actually works
O.K so the script has no option to detect option ft_psk_generate_local '1' so I crudely added append "$var" "ft_psk_generate_local=1" "$N" at line 688 where the commented out lines are.
However every time I change a wifi setting option nasid '5F83C411FD66' is deleted. I can be bothered to work out what is causing this, so I am just going to use the same hack and add append "$var" "nas_identifier=5F83C411FD66" "$N"