Max wait & fail fast

When you’d rather error than block for too long, cap the wait. A wait above the ceiling — counting every round of it, if the limiter has to wait again after losing a slot to another worker — throws RateLimitExceededException (extends RateLimitException) instead of sleeping:

Http::rateLimit(RateLimits::perHour(10)->maxWait(5_000)) // milliseconds
    ->get('https://api.example.com/report');

Handling the exception

The exception carries the owner key, the wait that would have been needed and your ceiling, so you can reschedule, degrade or report precisely:

use RoundlyConsulting\HttpClientRateLimits\Exceptions\RateLimitExceededException;
use RoundlyConsulting\HttpClientRateLimits\Facades\RateLimits;

try {
    $response = Http::rateLimit(RateLimits::perHour(10)->maxWait(5_000))
        ->get('https://api.example.com/report');
} catch (RateLimitExceededException $e) {
    $e->key;        // the owner key of the window that forced the wait
    $e->delayMs;    // the wait that would have been needed
    $e->maxWaitMs;  // your ceiling (5000)
}

The check runs before the RequestDeferred event and before any waiting, so nothing is recorded for the rejected call. In a named profile, set it with the max_wait key (milliseconds).