fix(sharding): drop stale cluster on join failure

Author: donhardman

src/Plugin/Sharding/Queue.php

@@ -192,43 +192,9 @@ protected function shouldSkipQuery(array $query): bool {
 			}
 		}
 
-		// Defer ALTER CLUSTER ... ADD until its target internal cluster has re-formed a primary
-		// view on this node. During a rejoin/failover rebalance the queue keeps processing even
-		// while internal clusters are non-primary (to avoid deadlock), but issuing the ADD before
-		// the returning member has re-synced is premature: the daemon reports "nodes not ready"
-		// and, in the worst case, crashes in the SST donor setup (SetDonor4Joiner). Waiting here —
-		// without burning a try — lets the bootstrap + Galera reconnect bring the cluster primary
-		// first. Primary is reached independently of this ADD, so this cannot deadlock.
-		if ($this->isAlterClusterAddTableQuery($query['query'])
-			&& !$this->isQueryClusterPrimary($query['query'])) {
-			Buddy::debugvv("Sharding queue: defer {$query['id']} — target cluster not primary yet");
-			return true;
-		}
-
 		return !$this->attemptToUpdateStatus($query, 'processing', 0);
 	}
 
-	/**
-	 * Whether the internal cluster targeted by an ALTER CLUSTER statement is primary on this node.
-	 * Unknown/unparseable cluster -> treat as ready (do not block).
-	 * @param string $query
-	 * @return bool
-	 */
-	protected function isQueryClusterPrimary(string $query): bool {
-		$name = $this->extractClusterNameFromQuery($query);
-		if ($name === null) {
-			return true;
-		}
-		try {
-			$res = $this->client->sendRequest("SHOW STATUS LIKE 'cluster_{$name}_status'")->getResult();
-			/** @var array{0?:array{data?:array<array{Value?:string}>}} $res */
-			$status = $res[0]['data'][0]['Value'] ?? 'primary';
-			return $status === 'primary';
-		} catch (\Throwable $e) {
-			return true;
-		}
-	}
-
 	/**
 	 * Helper to process the query from the queue
 	 * @param  Node   $node