[BUG] Current implementation of auto-tune switch doesn't work correctly

[sectroyer]

Describe the bug
I tried running a scan with auto-tune switch on a target that allows maximum of 300 req/s. With auto-tune feroxbuster was trying to do almost 4k req/s. As a result it was receiving tons of 429 and still decided it's good idea to "increase speed" I let it run for around 30 seconds (to not overburden the target ) but speed kept on oscillating around 3.5k req/s. At the same I had no way to "hint" the app about max rate as --rate-limit option doesn't work with --auto-tune. I attach a screen shot of my scan.

[epi052]

i recently reworked the auto-tune logic so that it's more aggressive about returning to a faster speed. It'll check every --timeout / 2 seconds whether the # of 429s/errors has increased or not and adjust based off that. auto-tune always tries to work itself back to an unlimited scan, which may not be the best solution. It doesn't have the concept of 'locking in' a rate (i.e. 300 req/s).

as a test, you could also try a longer span between rate adjustments by increasing the timeout, to see how it affects behavior. it would give it more time to 'see' the additional 429s

currently, --auto-tune and --rate-limit are mutually exclusive. would being able to use --rate-limit as a maximum value for --auto-tune address your issue?

ref: https://epi052.github.io/feroxbuster-docs/docs/examples/auto-tune/