When using IIsi ROM images in my IIx, I had some problems with hardware specific things. Rocket software got upset because the IIsi wasn't supported but the IIx was, and a couple other minor annoyances. Some software checks the ROM checksum, revision, and there's a type encoding in the ROM that AFAIK is meaningless, but back in the day some software referred to it rather than using gestalt information because gestalt could be fooled more easily than swapping ROMs.
Aside from that, the ROM detects the machine type at runtime, and how exactly a machine advertised what it was changed over time. The II/IIx/SE/30 used a combination of some bits in VIA registers and address maps (some address ranges repeated on some machines, not on others, so if the VIA bits were the same, you could disambiguate based on the address mappings), then later models expanded the bits in the VIA registers to be a little more scalable, eventually there was the addition of a machine ID register which showed up in the IO portion of the address space. Eventually all this gets propagated to the machine type gestalt value that higher level software usually keys off of.
BUT, the machine id isn't the only thing going on. At the end of the ROM is a declrom describing "Slot 0" (starting with 2nd rev of the Mac II ROM). For things like the IIci/IIsi, which included video that did not exist in any sense the SlotMgr knew about, this declrom could provide drivers, etc. for non-slot based integrated peripherals in a scalable manner (as opposed to the SE and earlier). So your IIci ROM includes a declrom claiming there is RBV video in your SE/30, when there isn't. The driver probably gets loaded and Opened, but the Open should fail when it finds no hardware there, and from then on it's ignored. The SE/30's video lives in slot E, so the SlotMgr finds it, its declrom, and subsequently its driver.
Anyway, that's all just some background. It doesn't really address the video initialization question, sorry.
Aside from that, the ROM detects the machine type at runtime, and how exactly a machine advertised what it was changed over time. The II/IIx/SE/30 used a combination of some bits in VIA registers and address maps (some address ranges repeated on some machines, not on others, so if the VIA bits were the same, you could disambiguate based on the address mappings), then later models expanded the bits in the VIA registers to be a little more scalable, eventually there was the addition of a machine ID register which showed up in the IO portion of the address space. Eventually all this gets propagated to the machine type gestalt value that higher level software usually keys off of.
BUT, the machine id isn't the only thing going on. At the end of the ROM is a declrom describing "Slot 0" (starting with 2nd rev of the Mac II ROM). For things like the IIci/IIsi, which included video that did not exist in any sense the SlotMgr knew about, this declrom could provide drivers, etc. for non-slot based integrated peripherals in a scalable manner (as opposed to the SE and earlier). So your IIci ROM includes a declrom claiming there is RBV video in your SE/30, when there isn't. The driver probably gets loaded and Opened, but the Open should fail when it finds no hardware there, and from then on it's ignored. The SE/30's video lives in slot E, so the SlotMgr finds it, its declrom, and subsequently its driver.
Anyway, that's all just some background. It doesn't really address the video initialization question, sorry.