2026-11-17
Residential Session Lifetimes and Interrupted Flow Recovery

Residential gateways can rotate exits or try to retain an exit for a session. The exact behaviour depends on the provider, pool and settings. This guide focuses on sizing a session lifetime, keeping task records useful and recovering an interrupted flow without confusing a temporary sticky session with a static IP allocation.

Two modes, one pool

Rotating session. Requests or connections are routed according to the pool rotation policy. Addresses may recur, and separate requests do not necessarily represent different visitors or independent outcomes.

Sticky session. A provider-defined identifier asks the gateway to retain an exit for a configured period. Check the actual syntax, lifetime limit and behaviour when an exit becomes unavailable; neither a timer nor an identifier guarantees continuity.

Both modes may use the same residential pool. Availability, address reuse and the provider’s routing rules still matter, and a sticky session does not necessarily reserve an exclusive address.

Sticky session is not a static IP

This is the most common source of confusion, so it deserves its own section.

A sticky residential session is temporary. An exit may become unavailable before the configured lifetime ends. Depending on the service, subsequent requests may fail or use another exit; detect this behaviour instead of assuming automatic reassignment.

A static residential or ISP proxy is sold as a longer-lived IP allocation, often hosted on different infrastructure from a rotating residential pool. Confirm the provider’s actual address-retention, replacement and exclusivity terms. A static allocation is not a guarantee against outages.

Rule of thumb: use a measured sticky lifetime when a short flow needs continuity. For a long-lived allowlist or a workflow that requires the same address over days, evaluate a static allocation and confirm its retention and replacement policy.

Choosing a session lifetime

Don't pick a lifetime by habit ("10 minutes is fine"). Estimate it from the flow you are running:

Initial lifetime estimate ≈ measured active flow duration + a tested margin, within the provider’s supported limits.

Illustrative example: a fare-check flow has a search, a results page, three pagination steps and a details request. Assume six sequential steps, each averaging 20 seconds including rendering and pacing.

6 × 20 s = 120 s

+ 50% margin = 180 s

A three-minute starting estimate allows a 50% margin in this example. Validate it against slower runs and the service’s supported session settings; it is not an availability guarantee.

A longer timer does not itself make a flow more reliable or create extra requests. Avoid reusing a session across unrelated jobs, release local state when the flow ends and separately control the request rate.

FlowTypical requestsIllustrative starting point; validate with measured duration
Single product or listing page1Rotating
Search + 2–5 result pages3–6Measure flow duration; 2–5 min can be a pilot setting
Fare/price check with filters and details5–10Measure flow duration; 5–10 min can be a pilot setting
Multi-step form you are permitted to use (quotes, availability)8–15Measure flow duration; 10–15 min can be a pilot setting
Same identity needed for hours or days—Evaluate static allocation and retention terms

Session-ID patterns that keep large jobs manageable

Assign a session to the logical flow when continuity matters. A worker can process many unrelated tasks, so avoid carrying cookies, session identifiers or credentials from one task into the next.

Use opaque, unique session IDs and log a separate internal job reference. Avoid putting personal data, account names or sensitive search details into proxy usernames or broadly accessible logs.

End local session state deliberately after the flow. A new identifier may change routing, but it does not guarantee a previously unused exit or release a provider-side resource unless that service supports an explicit release operation.

Cap both concurrent flows and aggregate requests per target. Twenty session identifiers do not prove twenty unique IPs. Measure observed exit reuse and honour target and gateway rate limits across workers.

Symptoms of the wrong mode

What you seePossible cause to investigateFix
Page 2 of results shows page 1 or a reset searchIP, cookies, parameters or server-side session state changedCompare request/state logs; use sticky routing if the flow requires it
Cart, filters or chosen currency "forget" themselvesSession-state, locale, account or routing mismatchKeep the required state consistent and verify the actual cause
Extra verification on every pageRate, shared-exit reputation, cookies or other verification signalsInspect the response; reduce aggregate load and follow retry guidance
Sudden 403/timeouts halfway through a long sticky sessionExit availability, access policy, rate limits or a network failureCheck the error and observed exit before deciding how to recover
Traffic use higher than expectedRetries replaying whole flowsRetry a safe/idempotent step only within a bounded policy

Handling a dropped sticky IP gracefully

Plan for interrupted exits and distinguish a transport failure from an unknown application outcome:

Record the observed exit at suitable checkpoints using an appropriate IP-check endpoint. An ordinary target response may not reveal your exit address, and each check also consumes traffic.

If the exit changes, check whether the application actually requires IP continuity and whether cookies and server-side state remain valid. Pause a dependent flow when its state is uncertain.

Restart only a flow that is safe to replay. For form submissions, purchases or other state changes, reconcile the result or use a supported idempotency mechanism before retrying; do not blindly repeat a step whose outcome is unknown.

Log the session ID, checkpoint, error category and observed exit change. Use those observations to adjust routing, lifetime and retry limits; shortening the timer alone may not resolve the cause.

Residential or mobile for sticky sessions?

Residential and mobile services may offer sticky routing, but continuity depends on the actual pool, gateway and plan. Test the behaviour you need rather than inferring it from the network label.

Residential exits use broadband networks. Compare measured continuity, supported locations and actual billing for the intended job instead of assuming they are always cheaper or more stable.

Mobile exits use carrier networks, often with shared public addresses. A carrier-network vantage point does not reproduce a phone’s hardware, browser, operating system or radio performance.

PSB Proxy offers residential and mobile proxy services. Compare the supported session settings, locations and billing model with a pilot flow, and use a mobile pool when carrier-network behaviour is part of the requirement.

A simple decision checklist

Independent safe reads? Consider rotation with an aggregate rate limit and bounded retry budget.

Dependent steps within minutes? Consider sticky routing sized from measured flow duration, while preserving application state.

Same address required over days? Evaluate a static allocation and its retention/replacement terms.

Interrupted sticky flow? Inspect exit and application state, then retry or restart only when replay is safe.

Traffic above plan? Measure whole-flow retries and resources transferred; do not replay state-changing steps just to reduce traffic.

Keep routing policy, application state and recovery logic separate. A measured lifetime and clear job records make failures easier to investigate; they do not remove the need for availability checks, rate limits and safe replay rules.