Private or shared proxies
A shared pool costs a fraction of a private one, and on short jobs that works. On long ones it fails almost every time.
The actual difference
A private address is spent by you alone. A shared one is divided among strangers, and so is the platform limit that applies to it.
The consequence: on a shared pool you never know how much of the limit is left at the moment your request arrives.
How it looks mid-run
The run starts cleanly, then a refusal arrives halfway through for no reason on your side: same pacing, same concurrency.
Someone else exhausted the limit before you did. Nothing in your own logs can explain it.
When shared is fine
A one-off job of a few thousand requests where a broken run costs nothing. Testing an idea, a rough sweep, reconnaissance.
Anything on a schedule, or anything that ends up in a client report, runs on private addresses.
Telling them apart
Ask the vendor directly whether the address is issued to anyone else. Second signal: a price far below the market almost always means a shared pool.
shared platform limit divided among unknown users private the limit is yours, behaviour is predictable