test: fix flaky LLMQ signing recovery and PoSe ban assertions

- feature_llmq_signing.py: bump_mocktime(2) was shorter than the
  daemon's 5-second signing session cleanup cadence (CLEANUP_INTERVAL
  in signing_shares.cpp), so recovery responsibility never rotated to
  the next member. Use bump_mocktime(10) to guarantee at least one full
  cleanup cycle, and extend wait_for_sigs timeout from 2s to 15s for
  propagation margin.

- feature_llmq_simplepose.py: replace bare `assert check_banned()`
  with `wait_until(check_banned, timeout=10)`. After mine_quorum()
  returns, PoSe ban state may not be visible via RPC immediately under
  CPU contention.

Author: thepastaclaw

test/functional/feature_llmq_signing.py

@@ -196,9 +196,11 @@ def assert_sigs_nochange(hasrecsigs, isconflicting1, isconflicting2, timeout):
             self.wait_until(lambda: mn.get_node(self).getconnectioncount() == self.llmq_size, timeout=10)
             mn.get_node(self).ping()
             self.wait_until(lambda: all('pingwait' not in peer for peer in mn.get_node(self).getpeerinfo()))
-            # Let 2 seconds pass so that the next node is used for recovery, which should succeed
-            self.bump_mocktime(2)
-            wait_for_sigs(True, False, True, 2)
+            # Advance mocktime past the daemon's 5-second signing session
+            # cleanup cadence so recovery responsibility rotates to the next
+            # member. Use 10s to guarantee at least one full cycle completes.
+            self.bump_mocktime(10)
+            wait_for_sigs(True, False, True, 15)
 
 if __name__ == '__main__':
     LLMQSigningTest().main()

test/functional/feature_llmq_simplepose.py

@@ -203,7 +203,7 @@ def test_banning(self, invalidate_proc, expected_connections=None):
                     if check_banned(self.nodes[0], mn):
                         break
 
-            assert check_banned(self.nodes[0], mn)
+            self.wait_until(lambda: check_banned(self.nodes[0], mn), timeout=10)
 
             if not went_offline:
                 # we do not include PoSe banned mns in quorums, so the next one should have 1 contributor less