I am waiting for btrfs again. Glad it waited for the weekend now! :-)

I ran snapper cleanup timeline, and five minutes later, I had a system lock up! System wouldn't boot anymore, so the Fedora 25 live USB was booted again. (dmesg: https://paste.fedoraproject.org/502654/31421714/)

The interesting lines (exactly 100 of them) are of this form:

[  105.305599] BTRFS warning (device dm-2): orphan qgroup relation 0x2e8->0x20000000000c8

So, it seems that snapper cleanup timeline started actually deleting stuff for the qgroups compliance, and that something went wrong there. I ended up pasting those log lines in a temporary file (/tmp/log), and constructed this one liner (that can probably be a lot shorter):

cat /tmp/log | awk '{ print $10 }' | awk -F - '{ print $1 }' | xargs -n1 -I {} bash -c 'sudo btrfs qgroup destroy "$(({}))" /mnt'

It takes all those warning lines, gets to the 0x2e8 part (which is the qgroup ID), and triggers a btrfs qgroup destroy on each of it.

I also started a btrfs scrub, which is slowly progressing its way through the SSD, with "0 errors", but with this line on the btrfs scrub status:

WARNING: invalid data on line 1 pos 385 state 6 (near "-14115|canceled:0|fi") at cmds-scrub.c:638

I'm currently just waiting for the scrub to finish. Seems to be a bit stuck at 40.72GiB.

EDIT: ended up disabling quota's, after some seconds the scrub continued, CPU usage dropped to normal, and everything seems fine now. Moral of the story: don't use quota's for now.