@juiceme there's plenty wrong with the J1's btrfs setup / version though.

As you allude to, a filesystem should not cave in on itself when you near capacity and/or put a load of small files on it, and then refuse to be able to clear that up because deleting files requires more journal space, and there's no space left. But at the same time perform file operations, just slowly, so that's it's not immediately clear what the problem with the system is.

I have had to manually delete caches and other things (looping through some xargs madness zeroing-then-deleting on loop until it worked) down to about 5gb allegedly-free on several occasions just in order to successfully run the balancer in order to get out of the "stuck" mode.

Obviously it was less effort to write a balancer daemon to run around tidying up than it was to backport or reconfigure btrfs to the J1 kernel. I imagine the whole debacle has soured them, especially as it only became obvious after people had had devices for over 6 months. Testing that kind of long-running gradually-worsening bug is a total nightmare so I imagine "ah screw it, let's use ext" was an appealing way out.