sync scripts clobber logs

[Paul Collins]

Overlapping runs of the sync scripts clobber the current log, even if the lock is not yet stale. (The scripts have always been written this way, at least to my recollection.)

We should probably switch around the lockfile logic and the logfile redirection. Pre-sync output can be sent to a common log file that is not clobbered.

This is might trickier than it sounds, because the redirection is done by the sync script as part of the process of detaching from the parent process and pipipng through annotate-output.

Another idea might be to use >> for the log file, although in terms of debuggability I think I prefer keeping each sync log per-sync and logging pre-lock output to a separate file.