Motivation

Both z_shieldcoinbase rpc-tests were in DISABLED_SCRIPTS pending fixes. This re-enables them so they run in CI (follow-up to #138's CI-green work).

What was wrong & the fix

wallet_z_shieldcoinbase.py

• Consensus branch mismatch. setup_nodes activated only NU5 at height 1, so zebra mined a coinbase committing to NU5's branch ID while zallet's params expected the latest NU's — zallet rejected the first block ("coinbase tx's claimed height doesn't match its consensus branch ID"). Now activates NU5/NU6/NU6.1/NU6.2 at height 1, matching the ZebraArgs default.
• Wallet not synced before account RPCs. Added sync_all() before z_getnewaccount, which returns "Wallet sync required" until the wallet has committed ≥1 block.
• Coinbase-maturity off-by-one in the expected-mature count.

wallet_z_shieldcoinbase_multi_taddr.py

• Same activation-height + wallet-sync fixes (here the symptom was z_getaddressforaccount failing with "chain height is unknown").
• Restart redesign. The old miner-switch logic assumed MAX_BLOCK_REORG_HEIGHT = 99 and that a restart drops the non-finalized tail. The real value is 1000, and zebra now restores non-finalized state from a periodic backup — so the truncate-wallet + data_requests sync task exits the wallet on a "no such tx" response that should be recoverable zallet#447 SQL workaround no longer held (history recovery exited trying to fetch transactions zebra no longer had). The test now keeps the pre-restart chain exactly one block long, waits for zebra to back up block 1, then restarts with miner=B and only extends the chain — no truncate, no orphaned wallet records, no #447 workaround.

⚠️ Known CI blocker (zaino) — affects wallet_z_shieldcoinbase.py only

wallet_z_shieldcoinbase_multi_taddr.py passes in CI (shard-2). wallet_z_shieldcoinbase.py (shard-1) currently fails in CI, but not because of the test logic — it's a zaino issue:

zaino_state::chain_index::finalised_state::db::v1::write_core: syncing finalised blocks 185..=243
zaino_state::chain_index: Sync loop iteration failed (N/10) ... ErrorFromSource(LmdbError(MapFull))
Sync loop failed 10 consecutive times, giving up
→ Chain indexer task exited → "Exiting Zallet because an ongoing task exited"
→ the test's wallet RPC disappears → wait_for_mature_coinbase_count fails at ~563s

zaino's chain-index LMDB hits MapFull while syncing its finalized state. This test is block-heavy (100+ maturity blocks plus per-sweep mining, ~280+ blocks total), so it crosses the limit; the lighter multi_taddr (~101 blocks) stays under it. The failure is deterministic — a re-run reproduced it identically.

This is independent of the test changes here (the test passes locally against a zaino with an adequate map). CI builds zaino from zingolabs/zaino@refs/heads/dev; the default map size claims 384 GB but is effectively not being applied in that build, so a few-hundred-block regtest chain fills it.

Unblock: fix the chain-index LMDB map size on zaino dev (or expose it so integration-tests can configure it). Once that lands, wallet_z_shieldcoinbase.py should pass with the fixes already in this PR. Tracking separately with the zaino team.

Test evidence

Both pass locally against zebrad 5.2.0 and zallet 0.1.0-alpha.3:

wallet_z_shieldcoinbase.py            -> All z_shieldcoinbase tests passed! / Tests successful
wallet_z_shieldcoinbase_multi_taddr.py -> PASSED (12.5 ZEC swept from 2 receivers) / Tests successful

In CI: multi_taddr ✅ passes; wallet_z_shieldcoinbase.py ❌ blocked by the zaino MapFull above.

AI disclosure

Used Claude Code (Opus 4.8) to diagnose the failures, implement the fixes, and draft this PR.

🤖 Generated with Claude Code