// Wait briefly for witness notification before retrying
        try {
            Thread.sleep(2000);  // 2 second grace period
        } catch (InterruptedException e) {
            Thread.currentThread().interrupt();
        }
    }
    
    // Proceed with normal error handling/retry
    super.handleConnectionLoss(error);
}
```

        // start
        AccessTimeoutTarget accessTimeoutTarget = null;
        TimeoutTask accessTimeoutTask = null;
        if (accessTimeout != null) {
            accessTimeoutTarget = new AccessTimeoutTarget(Thread.currentThread());
            accessTimeoutTask = TimeoutManager.getInstance().addTimeoutTarget(accessTimeoutTarget, accessTimeout, false);
        }

        try {

    try {
      transformResult = doTransform(localFunction, sourceResult);
    } catch (Throwable t) {
      restoreInterruptIfIsInterruptedException(t);
      // This exception is irrelevant in this thread, but useful for the client.
      setException(t);
      return;
    } finally {
      function = null;
    }

    /*

- [ ] Implement RDMA transport layer
- [ ] Add RDMA buffer registration
- [ ] Create RDMA read/write operations
- [ ] Implement RDMA credits management
- [ ] Add fallback to TCP when RDMA unavailable

#### 5.3 Integration Points
- Create new transport type alongside TCP
- Modify negotiation to include RDMA contexts
- Update read/write operations for RDMA

---

### Phase 6: Witness Protocol

 *
 * <p>This class supports an optional fairness policy for ordering waiting producer and consumer
 * threads. By default, this ordering is not guaranteed. However, a queue constructed with fairness
 * set to {@code true} grants threads access in FIFO order. Fairness generally decreases throughput
 * but reduces variability and avoids starvation.
 *

{* ../../docs_src/sql_databases/tutorial001_an_py310.py ln[14:18] hl[14:15,17:18] *}

Usar `check_same_thread=False` permite que o FastAPI use o mesmo banco de dados SQLite em diferentes threads. Isso é necessário, pois **uma única requisição** pode usar **mais de uma thread** (por exemplo, em dependências).

            final int threadId = t;
            executor.submit(() -> {
                try {
                    startLatch.await(); // Wait for all threads to be ready

                    for (int i = 0; i < operationsPerThread; i++) {
                        String message = String.format("Thread-%d-Message-%d", threadId, i);
                        byte[] plaintext = message.getBytes();

    // A happens-before edge between the writes of these fields and their reads exists, because
    // in order to read these fields, the corresponding write to incompleteOutputCount must have
    // been read.
    private boolean wasCancelled = false;
    private boolean shouldInterrupt = true;
    private final AtomicInteger incompleteOutputCount;

/**
 * Factory for [calls][Call], which can be used to send HTTP requests and read their responses.
 *
 * ## OkHttpClients Should Be Shared
 *
 * OkHttp performs best when you create a single `OkHttpClient` instance and reuse it for all of
 * your HTTP calls. This is because each client holds its own connection pool and thread pools.
 * Reusing connections and threads reduces latency and saves memory. Conversely, creating a client

 *
 * <p>When {@code post} is called, all registered subscribers for an event are run in sequence, so
 * subscribers should be reasonably quick. If an event may trigger an extended process (such as a
 * database load), spawn a thread or queue it for later. (For a convenient way to do this, use an
 * {@link AsyncEventBus}.)
 *
 * <h2>Subscriber Methods</h2>
 *
 * <p>Event subscriber methods must accept only one argument: the event.
 *