Useless advertise addr option when no direct conn to masters on consul start

#### Overview of the Issue
When I use the consul, I expect the advertise_addr or advertise_addr_ipv4 option. The address assigned to a particular IP will perform the following function:
If I have any problems with the Internet, then in combination with the Bind_addr option: "0.0.0.0" I can hope that even if my VPN does not start right away and will not create the address indicated in "advertise_addr" then when the Internet appears - The consul will be able to use this option to advertise themselves to other cluster participants without fixing directly on this address. Consul does this.
However, for some reason, the consul directly depends on the availability of a direct connection with the masters.
THOSE. The consul agent is not autonomous and will not work without a direct connection with the cluster when starting. What makes many of his features such as local DNS strongly limited, because We should have a connection with the masters, otherwise when restarting - we will not be able to launch the consul with its entuns and locally - nothing will work.

Our ultimate goal is to make the consul autonomous from masters at launch.
But we see that if when the server is restarted (not the Consul service), which is the most strange, **there will be no connection with the master servers, then the consul will not launch its endpoints and local DNS server**. What is stupid, because they are local to be autonomous and not depend on others.
At the same time, we observe this behavior only if we restart the server itself where the consul is located. If you disconnect VPN (according to which the Consul cluster is carried out) and restart the consul service - everything will be in order. I do not understand why it is like this and it looks like a bug and some cache files largely affect the efficiency of the consul.

See more:
https://developer.hashicorp.com/consul/docs/agent/config/cli-flags#_advertise
and
https://developer.hashicorp.com/consul/docs/agent/config/cli-flags#_bind
---

#### Reproduction Steps

Steps to reproduce this issue, eg:

1. Consul cluster with at least 3 masters and 1 alive agent separated by each other and communicating in the local or VPN network.
2. Test1 - You must turn off access to the network through which the consul communicates. Then restart the consul service. Then check the work of the local features of the consul (Endpoints and/or DNS)
3. Test2 - You must turn off access to the network through which the consul communicates. Then restart the whole server where consul agent was running. **Then check the work of the local features of the consul (Endpoints and/or DNS) - its not launched!**
-->

### Consul info for both Client and Server

Client info

```
Output from client 'consul info' command here

agent:
check_monitors = 0
check_ttls = 2
checks = 7
services = 7
build:
prerelease =
revision = 33e5727a
version = 1.20.2
version_metadata =
consul:
acl = enabled
known_servers = 3
server = false
runtime:
arch = amd64
cpu_count = 8
goroutines = 75
max_procs = 8
os = linux
version = go1.22.7
serf_lan:
coordinate_resets = 0
encrypted = true
event_queue = 0
event_time = 351
failed = 0
health_score = 0
intent_queue = 0
left = 0
member_time = 58820
members = 79
query_queue = 0
query_time = 4

```

```json
Client agent HCL config

{
"primary_datacenter": "fsn1",
"datacenter": "hel1",
"node_name": "784",
"bind_addr": "0.0.0.0",
"advertise_addr": "10.100.20.37",
"advertise_addr_ipv4": "10.100.20.37",
"client_addr": "127.0.0.1",
"ports": {
"dns": 8600,
"http": 8500,
"serf_lan": 8301,
"grpc": 8502
},
"domain": "consul",
"recursors": ["1.1.1.1", "8.8.8.8"],
"dns_config": {
"enable_truncate": true,
"only_passing": true
},
"retry_join": ["10.100.10.11", "10.100.10.12", "10.100.10.13"],
"rejoin_after_leave": true,
"data_dir": "/var/local/consul",
"log_file": "/var/log/consul-debug.log",
"enable_syslog": false,
"enable_debug": true,
"log_level": "debug",
"log_json": true,
"encrypt": "redacted",
"encrypt_verify_incoming": true,
"encrypt_verify_outgoing": true,
"leave_on_terminate": true,
"acl": {
"enabled": true,
"default_policy": "deny",
"enable_token_persistence": true,
"tokens": {
"agent": "redacted"
}
},
"tls": {
"defaults": {
"verify_incoming": true,
"verify_outgoing": true,
"ca_file": "/usr/local/etc/ssl/consul-agent-ca.pem"
},
"internal_rpc": {
"verify_server_hostname": true
}
},
"auto_encrypt": {
"tls": true
},
"limits": {
"rpc_rate": 10,
"rpc_max_burst": 100,
"rpc_client_timeout": "300s",
"rpc_max_conns_per_client": 50
},
"cache": {
"entry_fetch_max_burst": 3,
"entry_fetch_rate": 0.333
},
"disable_update_check": true,
"check_update_interval": "60m",
"enable_script_checks": true,
"gossip_lan": {
"probe_interval": "6s",
"probe_timeout": "3s",
"retransmit_mult": 3,
"suspicion_mult": 5,
"gossip_interval": "1s",
"gossip_nodes": 3
}
}

```

Server info

```
Output from server 'consul info' command here

agent:
check_monitors = 0
check_ttls = 0
checks = 4
services = 4
build:
prerelease =
revision = 33e5727a
version = 1.20.2
version_metadata =
consul:
acl = enabled
bootstrap = false
known_datacenters = 2
leader = true
leader_addr = 10.100.10.13:8300
server = true
raft:
applied_index = 12409662
commit_index = 12409662
fsm_pending = 0
last_contact = 0
last_log_index = 12409662
last_log_term = 1371
last_snapshot_index = 12398574
last_snapshot_term = 1371
latest_configuration = [{Suffrage:Voter ID:f48b1b28-0af1-876b-f949-9ba9ccedde51 Address:10.100.10.13:8300} {Suffrage:Voter ID:1fac6648-f334-90c0-9657-368c9626fe00 Address:10.100.10.11:8300} {Suffrage:Voter ID:d368c927-8d7a-fb00-3571-f53b693c9394 Address:10.100.10.12:8300}]
latest_configuration_index = 0
num_peers = 2
protocol_version = 3
protocol_version_max = 3
protocol_version_min = 0
snapshot_version_max = 1
snapshot_version_min = 0
state = Leader
term = 1371
runtime:
arch = amd64
cpu_count = 2
goroutines = 1091
max_procs = 2
os = linux
version = go1.22.7
serf_lan:
coordinate_resets = 0
encrypted = true
event_queue = 0
event_time = 351
failed = 1
health_score = 0
intent_queue = 0
left = 0
member_time = 58820
members = 80
query_queue = 0
query_time = 4
serf_wan:
coordinate_resets = 0
encrypted = true
event_queue = 0
event_time = 1
failed = 0
health_score = 0
intent_queue = 0
left = 0
member_time = 1913
members = 6
query_queue = 0
query_time = 1

```

```json
Server agent HCL config

{
"bootstrap_expect": 3,
"server": true,
"advertise_addr": "10.100.10.13",
"advertise_addr_wan": "10.44.10.3",
"client_addr": "127.0.0.1",
"datacenter": "hel1",
"primary_datacenter": "fsn1",
"data_dir": "/var/local/consul",
"domain": "consul",
"recursors": ["1.1.1.1", "8.8.8.8"],
"enable_script_checks": true,
"dns_config": {
"enable_truncate": true,
"only_passing": true
},
"performance": {
"raft_multiplier": 1
},
"ports": {
"dns": 53
},
"enable_syslog": true,
"encrypt": "secret",
"encrypt_verify_incoming": true,
"encrypt_verify_outgoing": true,
"leave_on_terminate": true,
"log_level": "info",
"rejoin_after_leave": true,
"retry_join": ["10.44.10.1", "10.44.10.2", "10.44.10.3"],
"retry_join_wan": ["10.44.0.2", "10.44.0.3", "10.44.0.4"],
"acl": {
"enabled": true,
"default_policy": "deny",
"down_policy": "extend-cache",
"enable_token_persistence": true,
"enable_token_replication": true
},
"cache": {
"entry_fetch_max_burst": 3,
"entry_fetch_rate": 0.333
},
"disable_update_check": true,
"check_update_interval": "10m",
"gossip_lan": {
"probe_interval": "6s",
"probe_timeout": "3s",
"retransmit_mult": 3,
"suspicion_mult": 5,
"gossip_interval": "1s",
"gossip_nodes": 3
},
"ui_config": {
"enabled": true
},
"connect": {
"enabled": true
}
}
```

### Operating system and Environment details

Run on ubuntu 24.04 LTS

The report names advertise_addr, advertise_addr_ipv4, bind_addr, retry_join, local endpoints, and DNS behavior; begin by running the two reproduction cases with access to the masters disabled. Compare restarting the service with restarting the host, and define done as local endpoints and DNS starting in both cases while the agent can later advertise and reconnect.

Written by the indexing model from the issue text.