Skip to content
← Field notes
FundamentalsJune 27, 2026· 7 min read

Sticky vs rotating sessions: one question decides it

When to hold one IP and when to rotate, the concurrency ceiling that quietly burns good IPs, and why rotating mid-login is the most common self-inflicted ban.

Trump Proxies · Network operations

AUTHPROXYCARRIERGET api.tmpx.io/api/rotate/… { "ok": true — refresh in progress }TRUMP PROXIES // ROTATION API

Most rotation confusion dissolves under a single question: does the target bind a server-side session to the IP it started on? If yes, hold the IP (sticky). If no, rotate. Everything else is tuning around that one rule.

When to rotate

Rotate when the work is stateless and broad — public scraping, price monitoring, SERP collection — where each request stands alone and IP diversity spreads your footprint. A rough starting cadence for public data is a fresh IP every 10–50 requests or every few minutes, tightening to as little as 1–3 requests with delays on hardened targets.

When to stay sticky

Stay sticky whenever state lives on the server: log in, then make authenticated requests; add to cart, then check out. Rotating mid-session throws away the IP the session was bound to, invalidates the cookie, and trips an "unusual activity" check. It's the most common self-inflicted ban we see, and it's entirely avoidable.

FIG · rotation timeline
Sticky sessions with deliberate rotationsONE LINE · ONE DAYsticky · account sessionsticky · next tasksticky · scraping run↻ rotate↻ rotateROTATE BETWEEN TASKS — NEVER UNDER A LOGGED-IN SESSION · REQUESTS HONOURED AT MOST EVERY 5 MIN
Hold one IP through each stateful task; rotate deliberately between tasks — never underneath a logged-in session.

How mobile rotation actually behaves

On a real mobile line, the IP changes carrier-side when the connection re-establishes — like a real handset picking up a new address as it roams — rather than being swapped per request by an orchestration layer. That makes every mobile port naturally "sticky until you rotate," which is exactly the right default for login and checkout flows. You trigger a rotation when you want a fresh IP; otherwise the line holds.

The pattern that works

  1. 01Discover URLs and collect public data on rotating IPs, at a human-plausible rate.
  2. 02Switch to a sticky IP per authenticated account, and hold it for the account's lifetime.
  3. 03Rotate only *between* tasks or accounts, and rotate your User-Agent and TLS fingerprint in lockstep with the IP — a fresh IP under a stale fingerprint fools no one.

Can I get a fresh IP on every single request with a mobile proxy?

Not the way a datacenter rotating gateway does. Mobile IPs change when the carrier session re-establishes, so the natural unit is "sticky until rotate," which is ideal for account work. For per-request stateless breadth, datacenter or residential rotating pools fit better.

How often should I rotate?

Only when there's a reason: you're switching accounts, you've hit a rate limit, or a challenge appeared. Rotating on a fixed timer "just in case" mostly discards good, warmed sessions.

Run it on real hardware

Ready to try real mobile proxies?

Dedicated real-SIM lines in the USA, Austria and Germany. 7-day trial, unlimited data, self-serve portal.