Upload leads · classify via SPF/PTR/HTTP · filter by rules · export with per-company caps · webhook on completion
Two layers: per-worker (how many leads each worker pod handles in parallel) and per-target-host (how aggressively we hit any one webmail / DNS resolver).
Caps work in flight inside a single worker process. Scale horizontally by adding replicas, not by raising these.
Total leads a worker handles in parallel across all queues.
Hard limit on DNS calls per minute per worker. Trades latency for cluster-stability.
Same idea for outbound HTTP probes.
A token bucket per remote host. Without this, the parallel fan-out can hammer a single webmail and get blocked.
Average rate per host.
How many tokens accumulate when a host is idle. Allows a brief surge for the fan-out's first wave.
Stricter (or looser) limits for specific hosts. Glob syntax: *.example.com.
| Host (glob) | Tokens / sec | Notes | |
|---|---|---|---|