Make fetch default to fail on errors

Fixes #23501

lib/ansible/modules/files/fetch.py

@@ -27,9 +27,7 @@ short_description: Fetches a file from remote nodes
 description:
      - This module works like M(copy), but in reverse. It is used for fetching
        files from remote machines and storing them locally in a file tree,
-       organized by hostname. Note that this module is written to transfer
-       log files that might not be present, so a missing remote file won't
-       be an error unless fail_on_missing is set to 'yes'.
+       organized by hostname.
 version_added: "0.2"
 options:
   src:
@@ -50,10 +48,13 @@ options:
   fail_on_missing:
     version_added: "1.1"
     description:
-      - When set to 'yes', the task will fail if the source file is missing.
+      - When set to 'yes', the task will fail if the remote file cannot be
+        read for any reason.  Prior to Ansible-2.4, setting this would only fail
+        if the source file was missing.
+      - The default was changed to "yes" in Ansible-2.4.
     required: false
     choices: [ "yes", "no" ]
-    default: "no"
+    default: "yes"
   validate_checksum:
     version_added: "1.4"
     description:
@@ -80,6 +81,11 @@ notes:
       depending on the file size can consume all available memory on the
       remote or local hosts causing a C(MemoryError). Due to this it is
       advisable to run this module without C(become) whenever possible.
+    - Prior to Ansible-2.4 this module would not fail if reading the remote
+      file was impossible unless fail_on_missing was set.  In Ansible-2.4+,
+      playbook authors are encouraged to use fail_when or ignore_errors to
+      get this ability.  They may also explicitly set fail_on_missing to False
+      to get the non-failing behaviour.
 '''
 
 EXAMPLES = '''