test(client): wall-clock deadline + 10s ceiling for input-roundtrip response waits

Second distinct flake in this suite today (gate run: MoveEast observed mid-move at the 5s window). The old accounting accrued elapsed only on empty polls — load-dependent effective window. Wall-clock deadline via Time.get_ticks_msec; ceiling 5s->10s (exit-on-arrival, costs nothing when healthy). Targeted suite 6/6.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

client/tests/test_input_roundtrip.gd

@@ -8,7 +8,10 @@ class_name TestInputRoundtrip
 extends GdUnitTestSuite
 
 const CONNECT_TIMEOUT: float = 3.0
-const RESPONSE_TIMEOUT: float = 5.0
+# Ceiling, not a sleep — the wait loop exits the moment the expected state
+# arrives. 5.0s flaked under load (gate run 2026-07-16: MoveEast observed
+# mid-move at timeout); 10.0s costs nothing when healthy.
+const RESPONSE_TIMEOUT: float = 10.0
 const MAX_PORT_ATTEMPTS: int = 5
 
 var _server_pid: int = -1
@@ -162,11 +165,14 @@ func _send_and_receive(
 	var send_err := _bridge.send_message(encoded)
 	assert_that(send_err).is_equal(OK)
 
-	# Wait for a snapshot that reflects the processed input.
+	# Wait for a snapshot that reflects the processed input. Wall-clock
+	# deadline: the old sleep-count accounting only accrued on EMPTY polls,
+	# so a streaming server made the effective window unbounded in theory and
+	# load-dependent in practice — measure real time instead.
 	var min_tick: int = last_tick + 2
 	var snapshot: Variant = null
-	var elapsed := 0.0
-	while elapsed < RESPONSE_TIMEOUT:
+	var deadline_ms: int = Time.get_ticks_msec() + int(RESPONSE_TIMEOUT * 1000.0)
+	while Time.get_ticks_msec() < deadline_ms:
 		_bridge.poll()
 		var msg := _bridge.poll_message()
 		if msg.size() > 0:
@@ -187,7 +193,6 @@ func _send_and_receive(
 				break
 			continue
 		await get_tree().create_timer(0.05).timeout
-		elapsed += 0.05
 
 	assert_that(snapshot).override_failure_message(
 		"no snapshot received within %.1fs after '%s'" % [RESPONSE_TIMEOUT, action_name]