Full disk issues (despite free disk space) with ext4+lvm: likely
     (based on having hit this myself multiple times and multiple reports
      in the discussion)
Corruption of individual files: unlikely
     (It seems internal drive crc is effective in catching those kinds
     of small-scale corruptions. There don't seem to be many reports despite
     lack of fs-level checksumming.)
Loss of individual files: unlikely
Loss of big chunk of the filesystem with ext4+lvm: very unlikely
     (I'm limiting this to data loss because of file system corruption.
      A disk dying is something else.)
 
Full disk issues (primarily metadata space exhaustion) with btrfs: unlikely
     This has become much better over the last few years, I don't think this is
     an important worry.
Corruption of individual files: none
Loss of individual files: unlikely
Loss of big chunk of the filesystem with btrfs: unlikely, but possible.

And that last one is the big unknown. For the sake of discussion,
let's say it happens 1e-5 / machine / year. This might still
result in a handful of cases over all Fedora users. If we stagger the
introduction of the new default, we might get half or one-third of such
cases. And btrfs might be improved upstream in response to our reports,
so we'd not just be delaying such failures, but reducing the total number
of occurrences.

After writing this out, I'm starting to think that we should just switch
the default, period. Some people will follow the default, others will
pick a different layout, and staggering will occur naturally.