Struct PollGuard 

pub struct PollGuard { /* private fields */ }

Per-poll-operation state machine that handles backoff, error classification, and consecutive-empty tracking.

Create one per (exchange, poll_type) — fills and quotes can fail independently and need separate backoff clocks.

Implementations

impl PollGuard

pub fn new(label: &'static str, config: &RunnerConfig) -> Self

Create a new guard with backoff config derived from RunnerConfig.

pub async fn execute<T, F, Fut>(&mut self, poll_fn: F) -> PollOutcome<T>

where T: HasItems, F: FnOnce() -> Fut, Fut: Future<Output = Result<T, ExchangeError>>,

Execute a poll with built-in backoff + error handling.

The skeleton:

1. Apply current backoff delay (if any).
2. Call poll_fn.
3. Classify the result and update internal state.
4. Return a clean PollOutcome so the caller only handles data.

pub fn looks_exhausted(&self, threshold: u32) -> bool

Check if the poll source looks exhausted.

Used by the runner for backtest exit detection: if delay == 0 && guard.looks_exhausted(3) { break; }