Hi Dan, Yes, the corrupted DB is the same every time. I modified the VFS to set a sw breakpoint when the offset corresponds to the corrupted zone. It looks like the data is "generated" by `pager_write_pagelist()`. Screenshot with the stack frame and data can be found in <https://1drv.ms/u/s!AkNyZuYZUo8xumhWdCsHHajf3udU?e=H6GD0e> I will continue to investigate, but if you have an idea where I should look please let me know. Rares.