reference youtube caption scrape vpn ratelimit 20260715T124849

---

name: VPN exit IPs are pre-burned for YouTube caption scraping; residential IP is the privileged lane description: Tested gluetun/Proton VPN exit for yt-dlp caption downloads: 429 on first request, vs hundreds of requests before throttle on residential IP. type: reference

---

During the CreatorTrack transcript-corpus scrape (yt-dlp auto-caption pulls for ~550 videos), tested routing through the Proton VPN exit already available via media-server's gluetun container, hoping a separate rate-limit bucket would speed up a YouTube-throttled scrape.

Result: the Proton exit got HTTP 429 on the very first caption request. YouTube's page-metadata endpoint worked fine (no bot-wall), but the caption-download endpoint specifically was already hot. By contrast, the residential home IP earns hundreds of caption fetches before being throttled at all.

Why: popular VPN exit IPs are shared across hundreds of users all hitting the same high-value endpoints (YouTube captions, etc.), so they arrive already rate-limited from someone else's traffic. A home residential IP is comparatively privileged precisely because it isn't a shared, previously-abused address.

How to apply: when a future bulk-scrape (YouTube or similar) hits per-IP throttling, don't reach for a shared/commercial VPN exit as the fix, it's very likely to be worse, not better. The correct lever is patience (long quiet windows let the limiter reset) or a dedicated non-shared egress IP, not a popular VPN endpoint. Confirmed empirically on gluetun/Proton in media-server; that container was left untouched (throwaway test container only).

[auto-memory session 60d132de-4659-485c-8044-427af0ccce7f, confidence 0.70, mode staged]