Although I have my own TI-92 and TI-92 Plus, neither of them contain a removable ROM module in the expansion port. However, reading about that expansion port led me to wonder what would happen in the scenarios:

1. Attempting to power on with the module unplugged
2. Pulling the module while the calculator is on (most risky, unless you've done this already and the calculator survived)
3. Swapping ROM modules containing different versions
4. Plugging an older ROM module into a later TI-92 with the ROM already soldered on
5. When removing a Plus module on a TI-92 with soldered ROM, are memory contents from before the Plus module retained?

I do not know what kind of errors or damage that these experiments may produce, but if someone with module-based TI-92s is willing to test and share the results of any of these scenarios, that would be awesome.

I own a TI-92 with soldered ROM & a Plus module, but I cannot perform any tests right now (it is stored elsewhere). I do know that there is an Exec string that makes it reboot in TI-92 mode (until the next reset), which I tested back when I had the calculator available, but I never tried removing the module. It resets the RAM when executed, although I am not sure if that is because it jumps to the memory reset routine in the TI-92 ROM or if it is because the RAM contents are incorrect & cause a reset.

I strongly discourage inserting or removing a module while the calculator is turned on, because they are not designed to be hot-plugged, so it could either cause a short circuit (which might fry something in either the calculator or the module) or cause data lines to be powered while the power lines are not connected (which could also fry things, primarily on the module).

Totally overlooked this in the age of hot-pluggable USB, but I'd agree it's a very real concern. Edited my post to reflect that.

I have a bunch of 92s and 92 Plus', with and without modules, I too have not tried hot-swapping modules, and I remember trying a few of these a few years ago.
From memory, I think this is what happens:
Using an older rom module boots fine with the older rom.
Using a rom module in a newer calc that already has a rom chip soldered on the board makes it boot off the module (I might need to re-check this because typing it out makes me feel like that's unlikely even though I think that's what happened)
Attempting to power on without the module just does nothing.

That's what I'd expect to occur on 92+ HW1, as opposed to earlier 92 & 92 II.

I went to dump the ROMs of my TI-92 with Plus module, & I dutifully removed the (main) batteries when removing the module (& thus still do not know what would happen if you pull it while powered), but I accidentally reinserted it with the batteries still in (& possibly powered on?). In case it has any effect, I had installed Fargo II & dumped the internal ROM using TiLP just before the insertion.

Anyway, the result of inserting the module with the batteries in (& again, possibly even powered on) is that it put a bunch of graphical garbage on the screen. (I may at this point have tried 2nd-Hand-On.) I immediately pulled the main batteries out & then reinserted them, resulting in it not booting at all (& definitely tried 2nd-Hand-On to no avail). Removing the module (with the batteries removed) let it boot from the internal ROM, but reinserting the module (again with the batteries removed) would not boot from the module. I was worried I had fried it, but as a last effort, I removed all batteries including the backup battery & spam-pressed ON, first with the module removed (to verify how long it took to clear RAM), & then with the module reinserted. Upon reinserting the batteries, it finally succeeded at booting from the module, so it seems my mistake did not cause permanent damage.

Just leaving all batteries out for about 30 seconds did not clear the internal RAM, while also spamming ON for ten seconds or so did & also sufficed to reset whatever had gone awry with the Plus module. I suspect there was garbage in the module RAM &/or registers that prevented booting.

EDIT: I could not find the page that had the Exec string to switch to TI-92 mode, but the program Nostalgia does so, from which I extracted the relevant code.
Code:

Heh, I'd have freaked a bit about having fried the module, too.

TILP's ROM dumper isn't usually known for corrupting the calculator's memory. However, unless you're using the development version where I fixed that long-standing issue, your dump of the internal ROM is corrupted, because of a silly bit of code which replaces the contents of the cert mem area by 0xFF if(tifiles_calc_is_ti9x(handle->model) && addr >= 0x10000 && addr < 0x12000). Not good on the 92

Disassembling from my own memory, looks like switching to 92 mode is a matter of switching to supervisor mode, setting ssp, and jumping to the boot code entry point ?