Node.js engineer for Amazon stock monitoring and checkout rate-limit tuning

I run a Node 22 service that watches Amazon product availability and places orders on customer accounts during restocks. I need someone who has operated against Amazon at scale to improve two specific things. 1. Stock detection. Today I poll the all-offers fragment anonymously through a proxy pool at about 450 ms per read. After roughly 85 minutes on one gateway range, CloudFront starts answering 503 on half the reads. I rotate exits and sessions and spread pollers, which keeps it usable. I want someone who knows other sanctioned or lower-cost signals, how the official Business Product Search API behaves in practice, how to size and spread a proxy pool by provider, and how to measure detection latency honestly. 2. Cart and checkout rate limits. Each account posts one cart add per second across its products, and an account that opens too many checkouts in a short window gets refused for a while. I want someone to review the pacing and backoff logic, tell me where I am leaving speed on the table and where I am about to get accounts flagged, and propose changes we can test against the harness before a drop.