tensor_stack.pop_back();
    auto op_id_it = tensor_tape.find(tensor_id);
    if (op_id_it == tensor_tape.end()) {
      continue;
    }
    int64_t op_id = op_id_it->second;
    auto op_it = op_tape->find(op_id);
    auto result_op_it = result.op_tape.find(op_id);
    if (op_id == -1 || op_it == op_tape->end() ||
        result_op_it != result.op_tape.end()) {
      continue;
    }

    }

    taskFaker.runTasks()
    assertEvents(
      "take plan 0",
      "plan 0 TCP connecting...",
    )

    taskFaker.advanceUntil(250.ms)
    assertEvents(
      "take plan 1",
      "plan 1 TCP connecting...",
    )

    taskFaker.advanceUntil(500.ms)
    assertEvents(
      "take plan 2",
      "plan 0 cancel",
      "plan 1 cancel",
    )

Pero aún así, la experiencia final no es la mejor 😞.

---

Esta sería la historia paralela equivalente de las hamburguesas 🍔.

Para un ejemplo más "real" de ésto, imagina un banco.

Hasta hace poco, la mayoría de los bancos tenían varios cajeros 👨‍💼👨‍💼👨‍💼👨‍💼 y una gran línea 🕙🕙🕙🕙🕙🕙🕙🕙.

Todos los cajeros haciendo todo el trabajo con un cliente tras otro 👨‍💼⏯.

!!! note "Nota"
    El comando `uvicorn main:app` se refiere a:

    * `main`: el archivo `main.py` (el "módulo" de Python).
    * `app`: el objeto creado dentro de `main.py` con la línea `app = FastAPI()`.
    * `--reload`: hace que el servidor se reinicie cada vez que cambia el código. Úsalo únicamente para desarrollo.

En el output, hay una línea que dice más o menos:

```hl_lines="4"

            RequestTrace trace,
            RepositorySystem resolver,
            RemoteRepositoryManager remoteRepositoryManager,
            List<RemoteRepository> repositories,
            ProjectBuildingRequest.RepositoryMerging repositoryMerging,
            ReactorModelPool modelPool,
            Map<String, Object> parentCache) {
        this.session = session;
        this.trace = trace;
        this.resolver = resolver;

* `idea-full` is a full blown IntelliJ IDEA Community Edition to be used in the plugin module.

These dependencies are quite large, so depending on the quality of your internet connection 
you might face timeouts getting them. In this case, you can increase timeout by specifying the following 
command line parameters on the first run: 
    
    ./gradlew -Dhttp.socketTimeout=60000 -Dhttp.connectionTimeout=60000

	AmbientIPv6 bool
}

// RepairConfig struct defines the Istio CNI race repair configuration
type RepairConfig struct {
	// Whether to enable CNI race repair
	Enabled bool

	// The node name that the CNI DaemonSet runs on
	NodeName string

	// Key and value for broken pod label
	LabelKey   string
	LabelValue string

	// Whether to fix race condition by repairing them
	RepairPods bool

    val result = dns.lookup("google.com")
    assertThat(result.size).isEqualTo(2)
    assertThat(result).contains(address("157.240.1.18"))
    assertThat(result).contains(address("2a03:2880:f029:11:face:b00c:0:2"))
    val request1 = server.takeRequest()
    assertThat(request1.method).isEqualTo("GET")
    val request2 = server.takeRequest()
    assertThat(request2.method).isEqualTo("GET")

    ],
    visibility = [
        "//tensorflow:internal",
    ],
    deps = [
        ":abstract_operation",
        ":c_api_unified_internal",
        "//tensorflow/c/experimental/gradients/tape:tape_operation",
        "//tensorflow/core/lib/llvm_rtti",
        "//tensorflow/core/platform:errors",
    ],
)

cc_library(
    name = "gradients_internal",
    srcs = [
        "gradients.cc",

Instead of that, by being an "asynchronous" system, once finished, the task can wait in line a little bit (some microseconds) for the computer / program to finish whatever it went to do, and then come back to take the results and continue working with them.