Wallet fixture holding a real CAT and NFT — the current replica cannot falsify what it is used to check

[MichaelTaylor3d]

Parent epic: https://github.com/DIG-Network/dig_ecosystem/issues/3166

Task

Obtain a wallet fixture that holds one unspent $DIG CAT coin and one NFT, so the $DIG-balance
family's acceptance bar can be measured instead of argued.

Why — the current replica cannot falsify the thing it is used to check

The node's live wallet is at %LOCALAPPDATA%\DigNode\wallet.sqlite (dig-node 0.160.0, synced to height
9,212,691). Measured 2026-08-28:

coins	948
distinct puzzle hashes	1
coins with asset_id set	0 — all NULL, i.e. all typed XCH
coins carrying a hint	0 of 948
cats / nfts / dids rows	0 / 0 / 0
derivations	0
unspent	2 coins, 1.599 XCH

Every table CAT/NFT/DID attribution populates is empty, and no coin has ever been hinted. CAT coins
live at cat_puzzle_hash(owner_p2, asset_id) — a different puzzle hash from the owner p2 — so a
node that has never run attribution has never subscribed to its own $DIG address and has never seen
those coins at all. That is #382 visible in production.

The consequence for testing is the point of this ticket. All 948 rows sit at the single owned p2
hash, so the point-read tier stages zero of them:

[REPLICA] nfts=0 dids=0
[ROUTE] point-read tier: believed=948 staged=0

A run against this replica proves no regression and corroborates nothing. nfts=0 dids=0 is
equally consistent with a defect having emptied them and with the wallet never having held an NFT — so
the measurement cannot distinguish the two, in either direction.

That has already cost real work. PR #393 reached eight review rounds; its round-7 gating finding
(the point-read narrowing terminally deleting nfts/dids rows) could only be demonstrated on a
simulator fixture, and its round-8 fix could only be verified there too. The earlier
{promoted: 8} → dig_balance 3856455 measurement came from capture files that are not on this machine,
and it predates findings it therefore cannot certify.

Without this fixture, every future round re-litigates the same unfalsifiable measurement.

What is needed

A wallet whose database, after a sync, contains at least:

1. One unspent $DIG CAT coin — so a balance can be wrong in the direction the family cares about,
   and so promotion has something real to promote.
2. One NFT or DID owned by the wallet — so the point-read tier stages a singleton and the
   nfts/dids path is exercised end to end rather than by fixture.

Either on the existing node wallet or a dedicated one. A dedicated test wallet is probably cleaner: it
can be minted deliberately, its expected contents are known exactly, and it does not entangle the
acceptance bar with whatever the main wallet happens to hold.

This is a mainnet spend and it is small. Note the current wallet holds 1.599 XCH unspent, so
funding is a real consideration rather than a formality.

Acceptance

A named wallet fixture, reachable by the dig-wallet harness (the DIG_REAL_WALLET /
DIG_REAL_COINS / DIG_REAL_PARENTS path), whose sync produces a non-zero cats row count and a
non-zero nfts or dids row count — so that a regression in either path makes the harness go red
rather than stay silently green.

Record the wallet's expected contents alongside it. A fixture whose correct output nobody wrote down
becomes unfalsifiable the moment its author stops working on it, which is how this ticket came to
exist.