Almost every mobile proxy listing promises rotation. Very few explain that the carrier decides whether you actually get a new address — and fewer still can tell you how often it works on their own line.
There are not enough IPv4 addresses for every phone in the world to have one. Mobile carriers solve this with carrier-grade NAT: your phone gets a private address on the carrier's internal network, and the carrier translates it to one of a smaller pool of public addresses shared among many subscribers.
So the address a website sees is not really "your" address. It belongs to the carrier and is shared. This is exactly why mobile IPs are hard for sites to block — and it is also why rotation behaves the way it does.
Rotating a mobile proxy means cycling the data connection so the carrier re-establishes it. The phone gets a fresh internal address, and typically a fresh public one too.
Typically. Not always. The carrier may reassign the same public address, particularly if the pool serving your cell is small or the network keeps a mapping for a while. The rotation succeeded; the visible result just did not change.
This is the difference between a vendor who has measured their line and one who has copied a feature list. "Instant rotation" is a claim about a mechanism. What you actually need to know is the change rate: out of a hundred rotations, how many produced a genuinely different public IP?
You do not need to take anyone's word for it, including mine. During a trial:
Also look at where the new addresses land. If every one sits inside the same /24 block, the pool is narrow and the addresses are visibly related — which matters if the destination looks at network neighbourhoods rather than single IPs. A vendor who has done this work will tell you the range; one who has not will change the subject.
rotations produced a different public IP
per rotation, end to end
Small sample, honestly stated, and growing. Every rotation is logged with the before and after address plus the elapsed time, and I will send the raw log on request. Rotations that return the same address are recorded as such rather than quietly counted as successes.
Want the raw rotation log, or to run the before-and-after test yourself on the live line? Ask and I will send it — including the rotations that returned the same address.