- Shipped
- August 5, 2026 at 6:02 PM UTC
- Author
- Kamo
- Commit
- 0ab6bf7
Hibernate creates the three new tables from the shared-lib entities; this adds what it cannot express — the partial unique index that keeps at most one open contact-point window per (lead, slot), and DESC covering indexes for the timeline's keyset scan. The seed opens a first window for every lead that predates the index. Not a communication back-fill: no historical call or message is associated to anything. Without it a pre-existing lead has no contact points at all and could never match an incoming call, so every tab would stay empty forever. It runs through LeadContactPointService rather than a bulk INSERT..SELECT so there is exactly one implementation of canonicalisation. A SQL twin would have to re-derive digit stripping, NANP handling, the extension floor and address unwrapping, and any later divergence would be invisible — seeded rows would simply stop matching live traffic. Applied against production: 66 contact points opened across 45 leads.