I found this link:
http://postgresql.1045698.n5.nabble.com/BUG-5507-missing-chunk-number-0-for-toast-value-XXXXX-in-pg-toast-XXXXX-td2126674.html And am also experiencing the same issue... More anecdotal evidence that this is a bug:
We recently(couple days ago) completely recreated the DB, and started inserting, so not data rot.
After a couple days of inserts, we could no longer read due to the TOAST error.
We had WAL shipping to a warm standby (we are using 9.0), and it also had the same issues after replaying the WAL.