AssumeUTXO: restart aborts with a divergent background chainstate #36495

test_snapshot_in_a_divergent_chain loads a UTXO snapshot while the background chainstate is on a stale lower-work branch below the snapshot base, then completes background validation without a restart. Add this before connecting node 3 to node 0:

@@ -256,6 +256,8 @@ class AssumeutxoTest(BitcoinTestFramework):
         assert_equal(normal['target'], target_str(REGTEST_TARGET))
         assert_equal(normal['target'], snapshot['target'])
 
+        self.restart_node(3)
+
         # Now lets sync the nodes and wait for the background validation to finish
         self.connect_nodes(0, 3)
         self.sync_blocks(nodes=(n0, n3))

The restart aborts in ReachedTarget() because startup assumes the background chainstate is already on the snapshot’s branch. A later restart with the same datadir aborts in the same way: restarting in this state should not abort.

Assertion failed: (!target_block || target_block->GetAncestor(m_chain.Height()) == m_chain.Tip()), function ReachedTarget, file validation.h, line 680.

This is a low-priority edge case that requires loading the snapshot while the node remains on the stale branch, then restarting before background validation catches up - unlikely, but the assertion over-promises regardless. I am documenting it as an example of the inconsistencies created by the secondary chainstate. Rather than adjusting each one separately, I think we should remove it as proposed in #32377.