fix: 重连成功后重置 outstandingPings 计数器，修复频繁重连问题

- Error closing connection 降级为 debug（TLS 关闭错误是断网预期行为）
- Write failed: batchBuffer is nil 降级为 debug
- Write operation failed 降级为 debug
- 避免长时间断网恢复时日志泛滥

Sources/Nats/NatsConnection.swift

@@ -356,6 +356,10 @@ class ConnectionHandler: ChannelInboundHandler {
         guard let channel = self.channel else {
             throw NatsError.ClientError.internalError("empty channel")
         }
+        
+        // 重连成功后重置 ping 计数器，避免累积的失败计数导致立即断开
+        self.outstandingPings.store(0, ordering: .relaxed)
+        
         // Schedule the task to send a PING periodically
         let pingInterval = TimeAmount.nanoseconds(Int64(self.pingInterval * 1_000_000_000))
         self.pingTask = channel.eventLoop.scheduleRepeatedTask(
@@ -901,7 +905,7 @@ class ConnectionHandler: ChannelInboundHandler {
             continuation.resume(throwing: errorToUse)
         }
 
-        let shouldHandleDisconnect = state.withLockedValue { $0 == .connected }
+        let shouldHandleDisconnect = state.withLockedValue { $0 != .closed && $0 != .disconnected }
         if shouldHandleDisconnect {
             handleDisconnect()
         }
@@ -921,19 +925,49 @@ class ConnectionHandler: ChannelInboundHandler {
         if let natsErr = error as? NatsErrorProtocol {
             self.fire(.error(natsErr))
         } else {
-            logger.error("unexpected error: \(error)")
-        }
-        let currentState = state.withLockedValue { $0 }
-        if currentState == .pending || currentState == .connecting {
-            handleDisconnect()
-        } else if currentState == .disconnected {
-            handleReconnect()
+            // 降级为 debug 级别，避免频繁输出错误日志
+            // uncleanShutdown 是常见的 TLS 关闭情况，不需要作为错误处理
+            logger.debug("Channel error (will reconnect if needed): \(error)")
         }
+        
+        // 注意：不在这里调用 handleDisconnect
+        // context.close() 会触发 channelInactive，由它负责处理断开逻辑
+        // 这样可以避免重复处理和过度重连
     }
 
     func handleDisconnect() {
+        // Prevent duplicate disconnect handling
+        let shouldProceed = state.withLockedValue { currentState -> Bool in
+            if currentState == .disconnected || currentState == .closed {
+                return false  // Already in disconnected/closed state
+            }
+            return true
+        }
+        
+        guard shouldProceed else {
+            return
+        }
+        
+        // Set state to disconnected after check
         state.withLockedValue { $0 = .disconnected }
         
+        // Clean up pending continuations to prevent leaks
+        if let continuation = serverInfoContinuation.withLockedValue({ cont in
+            let toResume = cont
+            cont = nil
+            return toResume
+        }) {
+            continuation.resume(throwing: NatsError.ClientError.connectionClosed)
+        }
+        
+        if let continuation = connectionEstablishedContinuation.withLockedValue({ cont in
+            let toResume = cont
+            cont = nil
+            return toResume
+        }) {
+            continuation.resume(throwing: NatsError.ClientError.connectionClosed)
+        }
+        
         // Safely clear batchBuffer first to avoid race conditions
         let bufferToRelease = self.batchBuffer
         self.batchBuffer = nil
@@ -957,18 +991,24 @@ class ConnectionHandler: ChannelInboundHandler {
                     try result.get()
                     self.fire(.disconnected)
                 } catch {
-                    logger.error("Error closing connection: \(error)")
+                    // 降级为 debug：网络断开时 TLS 无法完成正常关闭握手是预期行为
+                    logger.debug("Connection closed with error (will reconnect): \(error)")
                 }
                 // Only start reconnect after disconnect is complete
                 self.handleReconnect()
             }
         } else {
-            // No channel, start reconnect immediately
+            self.fire(.disconnected)
             handleReconnect()
         }
     }
 
     func handleReconnect() {
+        // Cancel any existing reconnect task to prevent multiple concurrent reconnections
+        if let oldTask = reconnectTask {
+            oldTask.cancel()
+        }
+        
         reconnectTask = Task {
             var connected = false
             while !Task.isCancelled
@@ -984,7 +1024,7 @@ class ConnectionHandler: ChannelInboundHandler {
                 } catch {
                     logger.debug("Could not reconnect: \(error)")
                     if !Task.isCancelled {
-                        try await Task.sleep(nanoseconds: self.reconnectWait)
+                        try? await Task.sleep(nanoseconds: self.reconnectWait)
                     }
                 }
             }
@@ -998,7 +1038,7 @@ class ConnectionHandler: ChannelInboundHandler {
             // If we got here without connecting and weren't cancelled, we hit max reconnects
             if !connected {
                 logger.error("Could not reconnect; maxReconnects exceeded")
-                try await self.close()
+                try? await self.close()
                 return
             }
 
@@ -1021,6 +1061,14 @@ class ConnectionHandler: ChannelInboundHandler {
 
     func write(operation: ClientOp) async throws {
         guard let buffer = self.batchBuffer else {
+            // If state is connected but batchBuffer is nil, this is a "fake connection" state
+            // Trigger reconnect to recover
+            let currentState = state.withLockedValue { $0 }
+            if currentState == .connected {
+                // 降级为 debug：这是断网恢复场景的正常状态
+                logger.debug("Write failed: batchBuffer is nil, triggering reconnect")
+                handleDisconnect()
+            }
             throw NatsError.ClientError.invalidConnection("not connected")
         }
         do {
@@ -1029,7 +1077,8 @@ class ConnectionHandler: ChannelInboundHandler {
             // Trigger reconnect on write failure - connection may be broken
             let currentState = state.withLockedValue { $0 }
             if currentState == .connected {
-                logger.error("Write operation failed, triggering reconnect: \(error)")
+                // 降级为 debug：网络断开时写入失败是预期行为
+                logger.debug("Write operation failed, triggering reconnect: \(error)")
                 handleDisconnect()
             }
             throw NatsError.ClientError.io(error)