docs(adr): let checked-in matches report without an operator

[patrick-hermann-sva]

Players are going to check in at their half of the table, by phone (NFC sticker/QR) or with a key fob at the zaehlwerk hub (#321, stuttgart-things/zaehlwerk#92, stuttgart-things/zaehlwerk-firmware#63). Once both have checked in, nobody else needs to stand at the table, but ADR-0015 still requires an operator for every zaehlwerk result. The maintainer decided that checked-in matches go without one. This writes that decision down before any code, as CLAUDE.md asks of anything that contradicts an ADR.

Changes

• ADR-0024 (German, like every ADR):
  • The two check-ins are the proof. Every check-in is created in Schmetterpause, never asserted by zaehlwerk.
  • POST /api/results sends both check-in ids instead of operator_id. Schmetterpause accepts the result only if both check-ins exist, belong to exactly home_id/away_id on different halves, have not carried a result yet, and have not expired. Each check-in carries exactly one result.
  • New entered_via = 'checkin'.
  • reported_by may be null, but only for checkin: check (reported_by is not null or entered_via = 'checkin').
  • The result stays pending, as ADR-0015 says for the test phase.
  • The Definition of Done shows these rows in their own column, as it does for scoreboard.
  • Rejected alternatives: the home player as reporter, a "Zählwerk" observer account, a bare checked_in: true flag, dropping the operator for all zaehlwerk matches.
• ADR-0015 status: "Der Operator ist Pflicht" is superseded for checked-in matches only. The kiosk (ADR-0014) is untouched.
• docs/adr/index.md regenerated with scripts/adr_index.sh, and the ADR added to the nav in mkdocs.yml.

Verified

• sh scripts/docs_test.sh → "ADR index and nav are current".
• Docs only, no Go touched, so go test was not run.
• The MkDocs site was not built locally.

Left open, deliberately

• Numbering. feat: RFID fobs as an identity — link a fob in a short window, resolve it for zaehlwerk #321 had planned 0024 for the fob identity. This decision came first and took 0024, so the fob ADR becomes 0025. The comment on feat: RFID fobs as an identity — link a fob in a short window, resolve it for zaehlwerk #321 is updated to match.
• Status accepted. The core (no operator once both have checked in) is the maintainer's call from the issue discussion. The mechanics are my proposal and are open to change in review: check-in ids in the result, a nullable reported_by instead of an observer account, a validity of "a few hours" as a code constant.
• Rests on invariant 8 (relaxing not null and extending a CHECK constraint touch no rows) and invariant 3 (check-ins in Postgres, not Redis).
• No migration, handler or zaehlwerk change here. Those follow in feat: RFID fobs as an identity — link a fob in a short window, resolve it for zaehlwerk #321 and zaehlwerk#92.

Refs #321

🤖 Generated with Claude Code

https://claude.ai/code/session_01BPzjRwDRcLSv7QBeECfkyS

---

Generated by Claude Code