de:wasserversorger-enrich

Journal updated. The run completed cleanly end-to-end.

Processed a 45-company batch (23+22 shards) for the wasserversorger-enrich campaign: all verified as genuine drinking-water suppliers, merged into enrichment.jsonl (2249→2294 lines, +45 distinct companies) and roster-links.jsonl (+232 links) with zero JSON errors or collisions. Population coverage holds at 99.99% (unchanged, expected); flagged five new standing items for a human/future run (a likely official_domain correction for Stadt Klingenberg, a Birstein entity conflict, a Rodenberg lab-operator question, an aquavilla/Hausach follow-up, and a new pyrbaum.de connection-hang host trap) — all logged in the journal, nothing sent or submitted anywhere.

build_enrich_batch_excl.mjs 45 enrich-20260906-1530 → 0 pending_domain (roster 3375, 2244 enriched entering, 1151 pending, 6 standing duplicate-mint pairs excluded, unchanged). Split shard A (23) / shard B (22), dispatched as two foreground top-level Agent calls (run_in_background: false) in one message, each explicitly told not to spawn sub-agents or background work. Both ran synchronously (18 min / 38 min wall-clock).

Verification before merge: shard A wrote 23/23 lines to enrichment.part-shardA-enrich-20260906-1530.jsonl, shard B 22/22 to enrichment.part-shardB-enrich-20260906-1530.jsonl, both 0 bad JSON lines, exact match against the two batch files (0 missing, 0 extra), 0 collisions with the existing 2244 records. Appended into enrichment.jsonl: 2249 → 2294 lines (2244 → 2289 distinct company_ids, +45). map_roster.mjs: 21786 → 22018 links (+232, matching the shards' 99+133 exactly). consolidate_roster.mjs: still 177 shards / 3375 distinct companies (unchanged, expected); roster-links.csv 18632 distinct after dedup (+207). coverage.mjs: population 99.99% → 99.99% (unchanged, expected), same 24 residual municipalities / 11,023 people, Hartmannsdorf-Reichenau still largest at 1,040.

Non-suppliers found: 0 — all 45 confirmed genuine, currently active drinking-water suppliers.

Field recovery (shard A, 23 companies): name/website_url 23/23; address/phone 21/23; short_name 15/23 found (7 not_published); address/phone opening hours 15/23 and 14/23; email 18/23; contact_form_url 6/23 (11 not_published); emergency_phone_number 16/23; supply_area_municipalities 22/23; handelsregister_nr mostly not_applicable (20/23). 12/23 multi_zone.

Field recovery (shard B, 22 companies): name/address/phone/website_url/supply_area_municipalities 22/22; short_name 19/22; email 20/22; opening hours 17/22; emergency_phone_number 19/22 (0 not_published — every negative earned by opening a labelled Notdienst/Kontakt page first); contact_form_url 9/22 found; handelsregister_nr 1/22 found, 21/22 not_applicable. Analysis locator 16/22 (73%) usable own-domain source, 13/22 (59%) multi_zone.

New standing items from this pass:

• wv-de-4293 (Stadt Klingenberg) — shard A found the real Impressum-named entity is "Stadtwerke Klingenberg Kommunalunternehmen (AöR)" on stadtwerke-klingenberg.de, not an Eigenbetrieb on stadt-klingenberg.de as the roster stated. stadtwerke-klingenberg.de geo-blocked this session ("nicht verfügbar in Ihrem Land"), so all fields are sourced from search-cache at medium/low confidence — needs a live re-fetch and an official_domain correction once reachable.
• wv-de-18839 (Birstein) — a third-party directory names a distinct "Zweckverband Gruppenwasserwerk Birstein", conflicting with the roster's Gemeinde-self-supply-plus-WVK- technical-ops picture; unresolved because birstein.de 403s (see host trap below).
• wv-de-1151 (Rodenberg / Feggendorf zone) — the lab Auftraggeber for the "self-supplied" Feggendorf zone is actually Wasserverband Nordschaumburg, not the SG itself; flagged for a human rather than silently resolved.
• aquavilla GmbH (St. Georgen) identified as the shared technical operator for both wv-de-18854 (Königsfeld) and wv-de-10511 (Hausach) — same 24/7 emergency number 0800 0861861 for both. Hausach's arrangement (Oct 2023) postdates aquavilla's published per-municipality analysis list, so no PDF exists there yet — worth a retry later.
• Domain-collision scan after merge (official_domain claimed by >1 company_id, per the identity rule's failure-mode #1) found 28 pairs/triples total, none new this pass — all predate 2026-09-06 and are already-documented VG/Verband shared-site cases (e.g. vgrd.de: wv-de-4403 Gemeindewerke Dudenhofen + wv-de-4404 VG-Werke Römerberg-Dudenhofen, both correctly flagged in their own notes as "expected shared-domain case, not a merge target"). No action needed, logged so a future collision-scan diff has a clean baseline.

New host/technique traps for the list: birstein.de and stadt-klingenberg.de both return 403 to WebFetch and curl with a browser UA — same class as the documented l.de trap (bot protection, not absence). pyrbaum.de is a new failure mode: TLS handshake succeeds but every HTTP GET hangs indefinitely (curl and WebFetch both) — different from l.de's fast 403 or bodensee-wasserversorgung.de's cert-chain failure; worth a retry with a longer timeout rather than concluding absence. assling.de sub-pages 404'd to direct WebFetch despite being indexed live by search — inconsistent, not previously documented, not yet a confirmed durable trap. Shard B caught a WebFetch hallucination (claimed 4 Creußener Gruppe members; primary Mitglieder page re-check confirmed the roster's original 2) and a near-miss on a wrong same-named Gemeinde (gemeinde-rohrdorf.de Baden-Württemberg vs. the correct Bayern/Rosenheim Rohrdorf, wv-de-18859) caught via postal-code mismatch before use.

No merge/export scripts run beyond the standard finish pipeline. Nothing sent, no forms submitted, no captchas touched.

For next run: continue with build_enrich_batch_excl.mjs 45 (real batch tag). Still 0 pending_domain, 1106 other pending (3375 roster total, 2289 distinct company_ids enriched, 6 excluded). All prior standing items (Wassertrüdingen/Hesselberg-Gruppe merger question, plus every item carried in the 2026-09-06 10:30/05:30/00:30 and earlier entries — Reinfeld-Land member-list gap, Gerolzhofen domain correction, Ruhpolding AdöR-naming, Emmering concession-operator, Bernau/Aschau boundary, Gedern list_found reclassification, Weiskirchen per-zone follow-up, Wackersberg/Arzbach, Weida-Land host-down, Tuntenhausen list_found, Lambrecht split, Altensteig/Ebhausen cross-supply, Rednitzhembach HRB, Strohgäu domain correction, Feldkirchen (Lkr. München), Horstmar concession, Aldenhoven/Niederzier gap, Küps/Kronach boundary, Langenargen/ZWUS, ZWV Westpfalz contact-domain, two dead roster URLs, Schwarzenbruck wrong-entity, Lechraingruppe/Hardhofgruppe, Bad Endbach HR exception, Fuchstal/Vilgertshofen, Neunkirchen a. Brand/Rosenbach, and the older long tail) remain open, plus the five new items above (Klingenberg domain correction, Birstein Zweckverband conflict, Rodenberg/Feggendorf lab Auftraggeber, aquavilla/Hausach retry, pyrbaum.de hang trap).