Keep in mind that a bot will be picking some permutation of its stock library of UA+metrics info, and generating truly-random values for other more continuously-valued parameters (e.g. timing between actions), to try to find a combination that satisfies a backend integrity-check.
A "try again" just means "you haven't succeeded yet." If that's all you get, you're getting zero bits of new information — so you can't do anything other than to assume it was your timing that looked weird, and keep trying. (And you might be dealing with even more noise, e.g. trying to have the bot calibrate itself toward a very low human-tuned request rate limit, where above-rate-limit responses look no different than integrity-fail "try again" responses.)
Suddenly getting a (maybe permanent) hard-fail, meanwhile, means that you said something the integrity-checker really didn't like.
Presuming you have a lot of IP addresses to send requests from, you can then do many experiments to bisect the difference between a hard-fail and soft-fail, and use that to blacklist values from your UA+metrics library. It's free entropy!
A "try again" just means "you haven't succeeded yet." If that's all you get, you're getting zero bits of new information — so you can't do anything other than to assume it was your timing that looked weird, and keep trying. (And you might be dealing with even more noise, e.g. trying to have the bot calibrate itself toward a very low human-tuned request rate limit, where above-rate-limit responses look no different than integrity-fail "try again" responses.)
Suddenly getting a (maybe permanent) hard-fail, meanwhile, means that you said something the integrity-checker really didn't like.
Presuming you have a lot of IP addresses to send requests from, you can then do many experiments to bisect the difference between a hard-fail and soft-fail, and use that to blacklist values from your UA+metrics library. It's free entropy!