I can confirm that no other scripts are scheduled between 8:13 and 8:20 AM and no API logs were generated during that time. Thus, it seems that upon the first run of anything the database was already locked. After this, it stays locked.
Others have reported this but we haven’t identified why. It behaves very much like you are seeing, where it almost immediately is locked but you aren’t really doing anything.
I’ve tested in numerous environments and I cannot reproduce this. I’m curious if you can try with 5.6.13, because we did identify a object pinning issue related to the database. That said, it wasn’t fixed because of the locking and only because of memory growth in really busy servers.