Struct AdaptiveLoop 

pub struct AdaptiveLoop { /* private fields */ }

Closed-loop wrapper: drives a RateLimiter from netforge EngineStats deltas. Call AdaptiveLoop::tick every batch.

Decision rules:

• tx_drops increased since the last tick → packets are being lost on the way out (NIC ring full, kernel back-pressure). Halve the target rate via AdaptiveRate::report_drops.
• tx_drops flat AND tx_packets increased → call AdaptiveRate::report_success; after 10 clean ticks the rate creeps back up by 5% of the configured ceiling.
• The applied rate is propagated to the wrapped RateLimiter so the per-batch consume path actually slows down.

This is the interlock between the observed loss signal and the enforced token bucket. Without it the bucket would stay pegged at the configured ceiling regardless of what the wire is doing.

Implementations

impl AdaptiveLoop

pub fn new(max_pps: u64) -> Self

Construct with max_pps as the ceiling. Initial enforced rate starts at half (max_pps / 2) per AdaptiveRate::new.

pub fn tick(&mut self, tx_packets: u64, tx_drops: u64) -> u64

Process a fresh stats snapshot and return the enforced rate.

First call seeds the baseline and returns the initial rate without classifying anything.

pub fn apply(&self, limiter: &mut RateLimiter)

Apply the loop’s current target to a RateLimiter. Call after every tick — cheap (one assignment) and safe.

pub fn current_pps(&self) -> u64

Current enforced rate.