Phoenix backup via HBase snapshots: capturing secondary indexes and SYSTEM.CATALOG in one consistent restore

We use Apache Phoenix (5.x) on HBase 2.x and want a documented backup strategy based on HBase snapshots. The goal is a single restore procedure that brings back user tables, their Phoenix secondary indexes, and the SYSTEM.* catalog tables together, so that a restored table is queryable with correct schema and usable indexes.

The uncertainty is around consistency boundaries. Phoenix indexes live as separate HBase tables, and the catalog holds the schema metadata, so a snapshot of only the data table seems insufficient. But taking several snapshots at slightly different times may leave the data table and its index tables out of sync if writes are in flight. It is also unclear whether clone_snapshot (fast, same cluster) or restore_snapshot plus ExportSnapshot (off-cluster copy) is the right choice for a disaster-recovery RPO, given that snapshots only capture data at snapshot time.

Questions:

• Is there a supported way to snapshot a Phoenix data table, all its index tables, and SYSTEM.CATALOG as one atomic/consistent unit, or must writes be paused (or replication used) to avoid skew between them?
• After a restore, what is the recommended verification that index tables match the data table — beyond COUNT(*) and hash aggregates — and can EXPLAIN be relied on to confirm indexes are actually being served?
• For cross-cluster recovery, is ExportSnapshot of all related tables (including SYSTEM.*) the documented path, and are there version-specific namespace or catalog DDL caveats to watch for?