nx_polite_crawl.nx

nx_polite_crawl.nx -- the POLITE CRAWL GOVERNOR (operator: researcher must "not be small sampled but doesnt trigger bot blowback"). The key synergy this encodes: BROAD coverage and POLITENESS are the SAME move -- you reach hundreds of sources by spreading THIN across many hosts (1-2 hits each, paced), never hammering one. The anti-pattern that triggers bot-blocking is small-sample-but- aggressive (one host 100x). The governor enforces broad-and-gentle: * per-host MIN INTERVAL (>= N ms between requests to the SAME host) + robots crawl-delay * EXPONENTIAL BACKOFF on 429/503 (capped) -- back off the moment a host pushes back * round-robin LONGEST-IDLE scheduler -- always fetch the host idle longest = spreads load * BREADTH metric (distinct hosts / total requests) -- high breadth = both large-sample AND polite Per-host state in parallel i64 arrays (struct-free): last_ms[], backoff_until_ms[], consec_429[]. All integer; deterministic given a clock. Composes nx_research_synth (corroborate what's fetched) + nx_resource_governor (host CPU politeness). LAWS: struct-free, integer-only. license_tier: ORIGINAL