More details regarding the crash during ASIC ID reading stage,
In order for it to work the production have to hold down the reset button before running the OMAPFlash programming command (using the console). It works ~ 4 out of 5 times when running it from the console and fails 9 out of 10 time if running it from production code (which is executing a sequence of test commands and that's one of the first). Seems like a timing issue.