Problem

In multi-rig Gas Town setups where each rig has its own Dolt database, convoy auto-dispatch is completely broken. Convoys are always hq-cv-* (stored in the HQ database), but tracked issues use rig-specific prefixes (e.g., ds-* in the dashboard database). The dependency resolution cannot cross this database boundary.

Symptoms

• gt convoy status hq-cv-xxx → "Progress: 0/0 completed" despite tracking 14 issues
• gt convoy launch hq-cv-xxx → "tracks no beads"
• bd dep list hq-cv-xxx -v → finds 14 raw "external-deps" but reports "has no dependencies"
• Daemon logs: "Convoy hq-cv-xxx: empty but within grace period"
• gt convoy stage <epic> works (resolves deps at stage time) but the created convoy can't be launched or tracked

Root Cause

Two code paths in internal/convoy/operations.go fail for cross-database deps:

1. getConvoyTrackedIssues() (line ~293): Calls store.GetDependenciesWithMetadata() on the HQ store, which returns external:ds:ds-xxx refs. The fallback fetchCrossRigBeadStatus() (line ~336) resolves these via subprocess bd show, but the convoy status/launch commands don't use this path — they see 0 tracked beads.

2. isIssueBlocked() (line ~193): Calls store.GetDependenciesWithMetadata() directly with NO cross-database fallback. Returns stale/empty metadata for cross-rig blocking deps, causing incorrect dispatch decisions.

The Catch-22

• hq-* prefixed issues → same DB as convoy ✓ → but rigForIssue() returns empty (can't route) ✗
• ds-* prefixed issues → rigForIssue() routes correctly ✓ → but convoy can't resolve deps ✗

Impact

Multi-rig towns cannot use convoy auto-dispatch at all. Manual gt sling per wave is required, defeating the purpose of the convoy system.

Related Issues

Consolidates: #2605, #2606, #2604, #2423. Prior closed instances: #2373, #992, #915.

Proposed Fix

The daemon already has map[string]beadsdk.Storage with all rig stores open (daemon.go:openBeadsStores()). The data is available — just not wired through to the convoy operations.

Create a multi-store resolver that:

1. Accepts the store map from the daemon
2. Resolves issue IDs by extracting prefix → looking up the correct store
3. Provides ResolveIssues(ids, stores) used by both getConvoyTrackedIssues() and isIssueBlocked()

This replaces the fragile subprocess fetchCrossRigBeadStatus() with direct store queries — faster, no external bd dependency, and consistently applied to ALL resolution paths.

Scope: ~200-250 lines across 3 files. Zero risk to single-DB setups (multi-store lookup only activates when local store returns empty).

Environment

• Gas Town version: main (4d35143)
• Beads: 0.59.0
• Setup: Multi-rig town with per-rig Dolt databases (dashboard=ds, nav_server=ns, etc.)
• Daemon running with convoy auto-feed enabled