Should the FBM Sniper
go to the cloud?

You have two completely different workloads, not one. Wallapop, Vinted, and OfferUp are lightweight HTTP poll-and-compare jobs. They run on any $5/mo VPS, fit serverless cron, and need no browser. Facebook Cars, Facebook Electronics, and Mercari launch a full headless Chromium (0.5-1 GB RAM each), require a logged-in cookie session, and fail fast on datacenter IPs. Treating these as "the same product" is the source of your strategic confusion.

The thing you find most painful about self-hosting is your moat. Hard onboarding, no in-app proxy config, user-supplied Facebook sessions -- these are real friction points. They are also the reason your ban rate is low, your infra cost is $0, and you hold no user credentials. The moment you host the Facebook snipers yourself, you inherit per-user residential proxy bills ($50-200+/mo per decent pool), a logged-in FB session that gets 2FA-checkpointed, and a Terms of Service liability for storing those cookies. The friction you hate is the business model.

The request-based monitors are nearly free to host and have none of that baggage. A single Hetzner CX22 (4-5 EUR/mo) or Railway starter plan (~$5/mo) can run all three pollers for your entire user base. No Facebook session needed. Proxy requirements are minimal -- many users run these with no proxy at all. The per-user cost is effectively rounding error, and there is no credential-custody risk. This is the correct scope for a cloud-hosted Lite tier.

AWS and Firebase are not meaningfully better than a generic VPS for this use case, and are more expensive. Lambda and Cloud Functions can handle the lightweight pollers but impose execution-time limits and cold starts. The Puppeteer monitors need always-on, high-memory compute with a real IP -- that means EC2 (t3.small ~$15/mo) or Cloud Run with persistent sessions, both of which cost 2-3x a comparable Hetzner or DigitalOcean box. Managed cloud adds billing complexity without adding value for a solo founder running a community product.

The hybrid path is low-risk to try. Your Electron app stays exactly as it is. The Lite cloud tier is a separate, simpler service: three HTTP pollers, one small VPS, a Whop entitlement check, and WebSocket or webhook delivery. You can ship it without touching the existing codebase. If it does not convert, you turn it off and your core product is unchanged. Estimated 12-month operator cost for the Lite tier: $60-180 all-in for a modest user base.