We've been trying to track down a boot crash that happens less than 1% of boots.
General Test Methodology: Boot from sdcard, login to linux, powerdown, and then cycle power.
Attached is a description of the issue as of a month ago.
Since then we found out almost by accident that changing the version of the ti-linux-firmwares used to build u-boot affected the occurrence of this issue. And that moving to the ti-linux-firmware 09.01.00.003 tag has seemingly removed all Linux boot crashes from our tests, over 100k boot cycles.
This is good news however leaves a lot of confusion and uncertainty in what changed in the firmwares to resolve this issue.
What I'm looking for input on is the random crash behavior that was improved by updating the sysfw version. It is rather concerning to us that a binary blob firmware was causing us so much grief and we don't why the different versions behave differently.