Comment 39 for bug 1273764

My workaorund is setting GRUB_RECORDFAIL_TIMEOUT in /etc/default/grub to the timeout I need.

This is a very strange behavior and probably not intended!

This is the logic behind it:
# in /etc/grub.d/00_header there's a function that calls grub-probe and if the result is "lvm" (my case) it ends setting the variable recordfail_broken=1 (as if there would have been a problem while booting)
# because of this later on, and if booting in efi, the variable GRUB_RECORDFAIL_TIMEOUT is being used to set the timeout and not GRUB_TIMEOUT (as I would expect).

This is not the intended behavior as described in https://help.ubuntu.com/community/Grub2

So, if you are booting from a lvm volume, GRUB_TIMEOUT won't be used.

What is the intention of this? why lvm => recordfail_broken=1 which implies every boot is "as if last one didn't complete properly"... makes little sense to me though look intentionally made that way.