Spring Cloud Zuul - which connection parameters work when?

According to this comment, Zuul offers the following kinds of configuration parameters:

1. zuul.host.* : these include e.g. max-per-route-connections and max-total-connections. There are only relevant when routes are explicitly declared, as follows:

---
zuul:
  host:
    max-per-route-connections: 500
    max-total-connections: 5000
  routes:
    my-service:
      path: /my-service/v1/**
      url: http://some-api/v1/

2. When using Ribbon together with Eureka, the serviceId.ribbon.* parameters should be used - or, more globally (AFAIK) ribbon.* e.g. ribbon.MaxConnectionsPerHost. These settings would apply for all Ribbon clients. So, if we have for example:

---
ribbon:
  MaxConnectionsPerHost: 10
  MaxTotalConnections: 100
  ReadTimeout: 3000

and serviceA is registered via Eureka, then Zuul should create a RibbonCommand for the request with the above settings. This is seen in org.springframework.cloud.netflix.zuul.filters.route.RibbonRoutingFilter#forward:

protected ClientHttpResponse forward(RibbonCommandContext context) throws Exception {
    Map<String, Object> info = this.helper.debug(context.getMethod(),
            context.getUri(), context.getHeaders(), context.getParams(),
            context.getRequestEntity());

        RibbonCommand command = this.ribbonCommandFactory.create(context);
    try {
        ClientHttpResponse response = command.execute();
        this.helper.appendDebug(info, response.getRawStatusCode(),
                response.getHeaders());
        return response;
    }
    catch (HystrixRuntimeException ex) {
        return handleException(info, ex);
    }

}

So far so good. Except, I can't validate this behaviour.

In my application, I've set the following parameters:

-Dspring.profiles.active=test
-Dspring.cloud.config.label=local
-Deureka.client.defaultZone.serviceUrl=http://localhost:8761/eureka
-Deureka.instance.hostname=${spring.application.name}
-Dzuul.host.max-total-connections=2
-Dzuul.host.max-per-route-connections=2
-Dribbon.MaxTotalConnections=2
-Dribbon.MaxConnectionsPerHost=2
-Dribbon.MaxTotalHttpConnections=2

as well as the following config:

hystrix:
  command:
    default:
      execution:
        isolation:
          thread:
            timeoutInMilliseconds: 150000
          strategy: SEMAPHORE
        timeout:
          enabled: false

Afterwards, when I make the first request against the given service, I see that this.ribbonCommandFactory.create(context) a new HttpClientConnectionManager is created (org.springframework.cloud.netflix.ribbon.apache.HttpClientRibbonConfiguration.ApacheHttpClientConfiguration#httpClientConnectionManager), which has maxTotalConnections = 2 and maxConnectionsPerHost = 2.

Afterwards, I set a thread breakpoint just before command.execute(). I fire off >2 requests, expecting everything from the third onwards to complain about not being able to acquire a semaphore for execution.

What gives? Is this setting being checked during execute? Is my description above even correct? Which settings really apply when ribbon is being used with serviceIds?

I know that Spring Cloud Zuul is in maintenance mode, and I would like to switch to Spring Cloud Gateway ASAP, but I would still like to understand this behaviour.