Re-pin checklist: strict class at txWalFlush=1, TIMESERIES COMPACTION_INTERVAL (user's decisions 2026-10-04)

[tae898]

Two re-pin changes the user decided on 2026-10-04 must be on the re-pin checklist (CAMPAIGN.md section 7, the only list of re-pin changes), so they are not forgotten.

1. Row 5, the strict durability class: adopt txWalFlush=1 for ArcadeDB's strict class at the re-pin, disclosed on the page. The condition row 5 set ("keeps 2 unless 1 and 2 measure as equivalent guarantees") is now answered by upstream's own documentation: 1 is "Safe against power loss. Recommended for production." and 2 has "No additional recovery value over 1 ... Benchmarks show no measurable performance difference either." (concepts/transactions.adoc), and the server's production mode defaults to 1. It is the setting matched by effect to SQLite's synchronous=FULL in WAL mode (one fdatasync per commit), not a flattering one.
2. New row 54, the TIMESERIES declaration: declare ArcadeDB's time-series types with COMPACTION_INTERVAL 1 HOURS at the re-pin IF upstream confirms it as the intended declaration (TIMESERIES: is COMPACTION_INTERVAL the intended declaration for an hourly aggregate, with SHARDS left at its default? ArcadeData/arcadedb#9166), disclosed on the page and taught in the bindings docs; laptop evidence (relative only): compacted 12h aggregate 8.3-10.1 ms to 0.62-1.39 ms, first call 632-1,064 ms to 7.8-11.1 ms, PROFILE 0 fast-path blocks to 16, same answer. Only ArcadeDB has the knob, so it is the user's call (taken) and is disclosed.