Thanks for the welcome! The tricky part was that the orphan rate stayed around 60% even after I tuned the pool's job limits, which told me the problem wasn't in the pool but in the node itself. From there I narrowed it down to an integer overflow in the node code. After deploying the fix on Oct 8, the orphans dropped right away. I'm planning a full write-up of the debugging process as one of my next posts. And thanks for the Threads tip, I'll check it out.