If they are close enough, just wade in, everyone else does. If you think it would be clearer to start your own thread, or the current one is already 10-20 posts deep, do that and cite a couple of relevant threads. For different STM32 and QSPI IC best to start new threads, or be very specific about what you're using/talking about.
If memory chip, clocks and pins are consistent existing loaders can be used. For slight variations they can also be patched/modified in hours rather than days.
Good examples are sparse. ST didn't use HAL/CUBE for some of them, and the examples that do are often necessarily bloated and dependent on interrupts, which isn't necessary. HAL_Delay() issues can be addressed elegantly without.
On QSPI vs OSPI, there might be some L4+ examples floating around, but yes pretty much just need to switch the BSP code, make sure the interface comes up and maps the memory at 0x90000000. Programming dual-quad devices has some challenges as erase/write can complete out-of-order but if you wait for both it all works. Best to test the living snot out of the thing outside of the external loader environment as very difficult to debug otherwise. You can use board level LED's, USART's etc to provide side-band diagnostic output. Figure several man-days to properly implement and test.
These loaders shouldn't be as hard as ST makes them, QA teams need to test the stuff more thoroughly and hold everyone in the chain accountable..
