de:wasserversorger-discover

Ran a 60-unit singleton discovery batch (two synchronous shards, fully in-turn per the skill's anti-background protocol): 41 new water-supplier companies minted, 21 municipality links reused existing companies, 0 unresolved. Population coverage rose from 98.51% to 98.75% (679 municipalities / 1.04M people residual), roster now at 2,959 companies. No new domain collisions introduced; the standing 27-collision backlog (plus one name-only duplicate pair not caught by domain checks) still awaits a human consolidation pass. Journal updated with full details and next-run pointers (657 singletons remain).

Followed the skill's synchronous-shard protocol: two general-purpose subagents dispatched with run_in_background: false and the mandatory "no Agent/Task tool" line in their prompts; both completed inline within this turn, neither recursed into background sub-subagents.

State on entry: picked up cleanly after the prior 17:59 run (200ef0→7070a4 chain), whose merge was already fully landed on disk (roster.jsonl/coverage.json timestamped 18:24, matching the journal's reported 98.51%/10204/739). No stale leftover batch files needed reconciling this time. Refreshed with coverage.mjs + build_gv_layer.mjs before building a new batch anyway, per the skill's "never trust leftover files" habit.

Batch: build_batch60.mjs — 60 singletons (residual_population descending, same-day-exclusion intact). Split 30/30 into two synchronous subagents, company_id ranges wv-de-19330–19359 (shard A) / wv-de-19360–19389 (shard B).

Shard A (30): 20 new mints (17 self_supplied, 2 private GmbH, 1 new Zweckverband), 10 reused existing companies, 0 unresolved. Caught a live domain-confusion near-miss before minting: "Röttenbach" is ambiguous — Röttenbach bei Erlangen (roettenbach-erh.de) vs. this batch's actual target, Röttenbach in Lkr. Roth (roettenbach.de, AGS 09576141) — verified via AGS before fetching, same risk class as the vghr.de/vg-hr.de trap. Also confirmed a genuine one-domain-two-entities case: Bosau→ZVO Energie GmbH (wv-de-357) is distinct from its parent Zweckverband (wv-de-4207) despite sharing zvo.com.

Shard B (30, 31 link rows): 21 new mints, 11 reused existing companies, 0 unresolved. Saalburg-Ebersdorf genuinely split across two Zweckverbände by Ortsteil (ZV WALO + ZWA Obere Saale) — two link rows for one AGS, correctly not deduped per the skill's overlap rule. 2 new pending_domain rows (Zweckverband Buchenwasserversorgung; ZV Kammelgruppe, partial). Worked around a cert-mismatch host (bert-horn.de) via WebSearch rather than forcing the fetch — added informally to the running host-trap awareness, not yet worth promoting to the skill's formal list since it was a one-off, not a repeat offender like the documented traps.

Validation before merge: both shards well-formed JSONL (20/20, 21/21 roster rows; 30/30, 31/31 link rows). AGS-coverage check against batch60-input.json: all 60 assigned AGS present, 0 missing; the one apparent "duplicate" (16075135, Saalburg-Ebersdorf) is the intentional two-row overlap above, not an error. Cross-shard + full-roster eTLD+1 domain-collision recheck run before consolidate_roster.mjs, per the skill's "recompute across every shard, not just the existing roster" rule: 27 standing pre-existing collisions (all pre-dating this run, none touching either new shard), 0 new collisions introduced.

Merge: map_roster.mjs — needs-review queue unchanged at 1195 (singleton AGS-seeded rows don't feed it, as expected). consolidate_roster.mjs: 2918 → 2959 distinct companies (+41 = 20+21, exact). coverage.mjs: municipalities claimed 10204 → 10264 (+60, exact — 0 unresolved across both shards). Population 98.51% → 98.75% (+0.24pp), residual 739 → 679 municipalities / 1.244M → 1.044M people. build_gv_layer.mjs: multi-member pool unchanged at 10 units/22 municipalities (still on hold), singletons 717 → 657 (-60, exact). Post-merge domain-collision recheck (roster.jsonl + companies.jsonl, eTLD+1): 27, confirming this run introduced none.

Collision-count note for whoever reconciles next: measured 27 this run and last run's entry said 29 — not a regression, just a different measurement: the standing duplicate-mint pair wv-de-7505/wv-de-1415 (both "Zweckverband zur Wasserversorgung der Wiesenbachgruppe", both official_domain: null) doesn't show up in an eTLD+1 domain-collision check at all since neither row has a domain to collide on — it's a name-collision, not a domain-collision, and was probably being counted separately before. It's still unmerged and still needs the same human consolidation pass as the other 5 standing duplicate-mint pairs (3925/8502, 1148/8515, 7562/10522, 7722/11041, 10013/11012 — all still present, all still domain-collisions).

Crossed 98.75% population coverage.

For next run: continue singleton batches by residual_population descending — 657 singletons remain, largest still ~6.6k (Schelklingen etc., same-day-excluded until the date rolls over). The 10-unit/22-municipality multi-member pool remains on hold pending the overdue human decision. _tasks/netzbetreiber-contact/german-water/ remains absent from this checkout (noted every run).