What is a rotating proxy?

A rotating proxy changes the address your requests appear to come from, either on every request or on a timer. You connect to one endpoint and the provider handles which exit node each request goes out through, so a hundred requests can arrive at a site from a hundred different addresses.

The alternative is a sticky proxy, which holds one address for a defined period. Both have their place, and picking the wrong one causes problems that look like blocking but are not.

When rotation is right

Rotation suits stateless work: fetching many unrelated pages where no request depends on the one before it. Product pages from a sitemap, search results across many queries, a crawl of independent URLs. Each request stands alone, so a different identity each time costs nothing and keeps any single address well under a rate limit.

When rotation breaks things

Rotation destroys sessions. If a site issues a cookie tied to the address that received it, and your next request arrives from a different country, the session is invalid. Symptoms:

• Logged out immediately after logging in successfully.
• A cart or a multi-step form losing state between steps.
• Pagination that resets to page one.
• A 401 partway through an authenticated crawl.

These read as blocking. They are configuration. The fix is a sticky address for the duration of the session, not better proxies.

How rotation actually happens

Providers implement it two ways, and the difference matters for debugging. Per request gives a new address every time, with no control over which. Timed sessions hold an address for a set duration, typically between one minute and half an hour, after which it changes.

Rotating too aggressively is its own tell. A visitor whose address changes on every single request, while carrying one unchanged browser fingerprint, describes something no real person does. Many addresses sharing one identical and unusual machine is a pattern anti-bot systems look for directly.

Choosing between them

Ask whether request N depends on request N minus one. If it does, use a sticky address for the whole sequence. If it does not, rotate. Most crawls are a mix: rotate freely for the discovery pass, hold an address for anything behind a login.

Where Zenrows fits

Rotation is the default on Premium Proxy, so stateless scraping needs no configuration. Where you need the opposite, the session ID parameter pins requests carrying the same ID to one address, which is what keeps an authenticated sequence intact. Choosing per request rather than per account means one pipeline can do both.