I have had a endless running btrfs balance with the kernel
4.12.14-lp151.28.48-default. The balance was triggered from the
btrfsmaintenance balance timer.
I have no kernel stack trace, only the same following trace during 15h:
"kernel: [.......] BTRFS info (device dm-0): found 2087 extents"

The btrfs filesystem was in the following situation:
"
Total devices 1 FS bytes used 30.81GiB
        devid    1 size 45.00GiB used 45.00GiB path /dev/mapper/system-root2
"
I've tried to free some space through deletion of old snaphots during this
situation, but the btrfs balance was running without change.
Also canceling of the btrfs balance had no effect.
After these tries i have rebooted the system with skip_balance option in the
fstab.
Manually started btrfs balance in several steps (with argument -musage
beginning from 10 to 80, then with argument -dusage from 10 to 80)  were
running without problem, but the used size (second line) was a little smaller
than 45.00 GiB at the start of the balance runs.
A scrub of this filesystem shows no errors.
There are some other bug reports describing the same or a similar problem (e.g.
https://bugzilla.opensuse.org/show_bug.cgi?id=1160947), but for me it is not
clear, if this kernel, which has quite a lot btrfs related changes, should have
a bugfix for this problem.