Hi Greg O.
Our logs show the initial delete hit an out-of-memory error before it could commit. Moving to Standard raised the memory limit but kept the same duckling running, so the retries failed on leftover state from that first attempt rather than on memory. It went through once you landed on a fresh duckling about an hour later.
Going forward: and unqualified DELETE FROM holds deletion bookkeeping for every row in a single transaction, so chunking it with a commit per chunk, like you did in your later DML queries, keeps memory use flat not matter how large the table gets. If a write does fail, SHUTDOWN cycles the duckling before you retry, especially if you've upgraded the duckling size in between. docs here
.png){kind=logo}
