Logstash information:

Logstash version (e.g. bin/logstash --version): 8.17.6

Description of the problem including expected versus actual behavior:
Since the version 8.17.3, HTTP 404 error responses are handled as an empty pipeline list when using CPM to manage logstash configuration (see #17052).

It seems the new logic is only checking for HTTP 404 response codes to determine whether the managed pipeline list is empty, making it unreliable as HTTP 404 is a generic error and might happen in other situation as well, such as unknown Elasticsearch nodes (due to outages, for example). When that happens, pipelines are shutdown.

A possible solution would be improving the empty result detection maybe by also checking if the body is empty {} (returned by an Elasticsearch instance), or if it's another type of error, e.g. {"ok":false,"message":"Unknown resource."} returned by the gateway/API.

Steps to reproduce:

1. Configure LS with managed configuration
2. Start Logstash
3. Turn the node down or simulate an 404 response