* exists but cannot be used for another exchange, it is closed and this returns null.
   */
  private fun planReuseCallConnection(): ReusePlan? {
    // This may be mutated by releaseConnectionNoEvents()!
    val candidate = call.connection ?: return null

    // Make sure this connection is healthy & eligible for new exchanges. If it's no longer needed
    // then we're on the hook to close it.

 *  Fix: Improve recovery from failed routes.
 *  Fix: Accommodate tunneling proxies that close the connection after an auth
    challenge.
 *  Fix: Use the proxy authenticator when authenticating HTTP proxies. This
    regression was introduced in OkHttp 3.0.
 *  Fix: Fail fast if network interceptors transform the response body such that
    closing it doesn't also close the underlying stream. We had a bug where

    SerializableTester.reserializeAndAssert(emptyRangeMap);

    ImmutableRangeMap<Integer, Integer> nonEmptyRangeMap =
        new ImmutableRangeMap.Builder<Integer, Integer>()
            .put(Range.closed(2, 4), 5)
            .put(Range.open(6, 7), 3)
            .put(Range.closedOpen(8, 10), 4)
            .put(Range.openClosed(15, 17), 2)
            .build();

   * due to any of:
   *
   *  * A stale connection. The request was made on a reused connection and that reused connection
   *    has since been closed by the server.
   *  * A client timeout (HTTP 408).
   *  * A authorization challenge (HTTP 401 and 407) that is satisfied by the [Authenticator].
   *  * A retryable server failure (HTTP 503 with a `Retry-After: 0` response header).

    non-ASCII characters (including international characters and emoji).

 *  **Call canceling is more reliable.**  We had a bug where a socket being
     connected wasn't being closed when the application used `Call.cancel()`.

 *  **Changing a HttpUrl’s scheme now tracks the default port.** We had a bug
    where changing a URL from `http` to `https` would leave it on port 80.

          throw InterruptedIOException()
        }
      }
    }

  companion object {
    /** A large response body. Smaller bodies might successfully read after the socket is closed!  */
    private val BIG_ENOUGH_BODY = repeat('a', 64 * 1024)
  }

    operation ->> dep_func: Return from path operation
    Note over dep_func: Run code after yield
    Note over dep_func: ✅ Dependency closed
    dep_func ->> client: Send response to client
    Note over client: Response sent
    Note over dep_req: Run code after yield
    Note over dep_req: ✅ Dependency closed
```

	}
}

func (m *muxServer) unblockSend(seq uint32) {
	if !m.checkSeq(seq) {
		return
	}
	m.recvMu.Lock()
	defer m.recvMu.Unlock()
	if m.outBlock == nil {
		// Closed
		return
	}
	select {
	case m.outBlock <- struct{}{}:
	default:
		gridLogIf(m.ctx, errors.New("output unblocked overflow"))
	}
}

func (m *muxServer) ping(seq uint32) pongMsg {

For example, if you had a database session in a dependency with `yield`, the `StreamingResponse` would not be able to use that session while streaming data because the session would have already been closed in the exit code after `yield`.

This behavior was reverted in 0.118.0, to make the exit code after `yield` be executed after the response is sent.

/// info

		cacheStore:       cacheStore,
		namer:            namer,
		initialized:      make(chan struct{}),
	}
	// When the schema initialization is completed, the channel will be closed
	defer close(schema.initialized)

	// Load exist schema cache, return if exists
	if v, ok := cacheStore.Load(schemaCacheKey); ok {
		s := v.(*Schema)
		// Wait for the initialization of other goroutines to complete