This is the #1 question that I'm going to add. I just want to go over Entropy's notes and see what he's noticed already and then sum it up so we can hopefully point in a direction without telling a long story. I'm still waiting for a response from Samsung and I left a VM for the person at Sprint - I'll be passing this on to them as well. (From Sprint's angle using the GB code is a good solution if they want to go OTA - patching the eMMC would probably be better in the long term if possible.)
I'll give it until tomorrow around 9 am CDT in case anyone else has some lingering questions - then send it off. Figured it was better to make sure everyone gets a chance to catch up after the weekend before doing so.
What I've been able to figure out so far:
The main difference between I9100 update4 source and SHW-M250S Update5 source is that in I9100 update4, MMC_CAP_ERASE is not enabled in either the MSHCI drivers or SDHCI drivers. The MSHCI driver is the driver for our internal storage.
Even before Mr. Sumrall's responses, based on the fact this usually happened during a wipe/format, I suspected it was MMC_CAP_ERASE.
However, in Gingerbread kernels for I777/I9100, the MSHCI driver also has MMC_CAP_ERASE enabled. The driver does appear to be heavily customized with Samsung code, however. So I'm kind of baffled as to why MMC_CAP_ERASE seems safe on Gingerbread - Either the customization in the driver works around fwrev 0x19's bug, OR there's something elsewhere in the kernel that causes ERASE commands not to get fired at the chip even though it's enabled in the MSHCI driver.
The fact that Gingerbread was safe caused me to move away from MMC_CAP_ERASE until Mr. Sumrall's response regarding the ERASE bug in 0x19.
Either way, I believe that removing the MMC_CAP_ERASE from the MSHCI driver will render it safe. I'm just curious as to why Gingerbread isn't dangerous despite having this feature flagged as enabled in the MSHCI driver.
What I know for sure:
GT-I9100 Update3 - Gingerbread - MMC_CAP_ERASE enabled but still somehow safe
GT-I9100 Update4 - ICS - MMC_CAP_ERASE disabled and safe
SHW-M250S Update5 - ICS - MMC_CAP_ERASE enabled and dangerous
What we really need at this point is a way for the JTAG services to do a full reset of the chip - Even if they can't update the firmware past 0x19, as long as they can recover the chip to a normal working state, then people can have damaged devices repaired for $30-50 instead of $200+.