what is concurrency?
how many proxy connections you run at the same time, which sets both your throughput and most of your block risk.
bandwidth products rarely cap threads, but the pool, the target and your own machine all impose real limits. per exit, concurrency is what rate limits notice: fifty parallel requests through one address is a burst, the same fifty spread over fifty addresses is unremarkable traffic. so the number worth tuning is requests per second per exit, not the global total. exit type sets the ceiling too. residential and mobile exits are ordinary home and radio links, so each handles far fewer parallel streams than a server, and pushing harder buys timeouts rather than speed. count the client side as well, since each headless browser instance costs hundreds of megabytes and file descriptor, dns and cpu limits usually bite before the network does. start low, raise until error rates move, then back off. myx datacenter and isp take the highest concurrency; on residential and mobile, add exits rather than threads per exit.