100.100.100.100:53 no associated peer node

### What is the issue?

This evening I could no longer ping or ssh to a machine outside of my local network:

log output from trying a ssh session from 100.95.64.57 to 100.111.81.66:
```
default 00:38:47.227767-0400 IPNExtension Accept: TCP{100.95.64.57:52443 > 100.111.81.66:22} 64 ok out
default 00:38:48.231806-0400 IPNExtension Accept: TCP{100.95.64.57:52443 > 100.111.81.66:22} 64 ok out
default 00:38:50.236366-0400 IPNExtension Accept: TCP{100.95.64.57:52443 > 100.111.81.66:22} 64 ok out
default 00:38:52.231447-0400 IPNExtension open-conn-track: timeout opening (TCP 100.95.64.57:52443 => 100.111.81.66:22) to node [Q9inX]; online=yes, lastRecv=never
default 00:38:53.232903-0400 IPNExtension open-conn-track: timeout opening (TCP 100.95.64.57:52443 => 100.111.81.66:22) to node [Q9inX]; online=yes, lastRecv=never
default 00:38:55.237983-0400 IPNExtension open-conn-track: timeout opening (TCP 100.95.64.57:52443 => 100.111.81.66:22) to node [Q9inX]; online=yes, lastRecv=never
default 00:38:59.249519-0400 IPNExtension open-conn-track: timeout opening (TCP 100.95.64.57:52443 => 100.111.81.66:22) to node [Q9inX]; online=yes, lastRecv=never
```

ping from 100.95.64.57 to 100.111.81.66:
```
PING rafa-vps.rafaelgarrido.gmail.com.beta.tailscale.net (100.111.81.66): 56 data bytes
Request timeout for icmp_seq 0
Request timeout for icmp_seq 1
Request timeout for icmp_seq 2
Request timeout for icmp_seq 3
Request timeout for icmp_seq 4
Request timeout for icmp_seq 5
```
```
default 00:40:21.223790-0400 IPNExtension Accept: ICMPv4{100.95.64.57:0 > 100.111.81.66:0} 84 ok out
default 00:40:22.225975-0400 IPNExtension Accept: ICMPv4{100.95.64.57:0 > 100.111.81.66:0} 84 ok out
default 00:40:23.230720-0400 IPNExtension Accept: ICMPv4{100.95.64.57:0 > 100.111.81.66:0} 84 ok out
default 00:40:31.254240-0400 IPNExtension Accept: ICMPv4{100.95.64.57:0 > 100.111.81.66:0} 84 ok out
default 00:40:41.283225-0400 IPNExtension Accept: ICMPv4{100.95.64.57:0 > 100.111.81.66:0} 84 ok out
```

macos: tailscale debug --prefs (redacted keys)
```
tailscale debug --prefs
{
"ControlURL": "https://controlplane.tailscale.com",
"RouteAll": true,
"AllowSingleHosts": true,
"ExitNodeID": "",
"ExitNodeIP": "",
"ExitNodeAllowLANAccess": false,
"CorpDNS": true,
"WantRunning": true,
"LoggedOut": false,
"ShieldsUp": false,
"AdvertiseTags": null,
"Hostname": "",
"NotepadURLs": false,
"AdvertiseRoutes": [],
"NoSNAT": false,
"NetfilterMode": 0,
"Config": {
"Provider": "google",
"LoginName": "rafaelgarrido@gmail.com"
}
}
```

manjaro linux: tailscale debug --prefs (redacted keys)
```
{
"ControlURL": "https://controlplane.tailscale.com",
"RouteAll": false,
"AllowSingleHosts": true,
"ExitNodeID": "",
"ExitNodeIP": "",
"ExitNodeAllowLANAccess": false,
"CorpDNS": true,
"WantRunning": true,
"LoggedOut": false,
"ShieldsUp": false,
"AdvertiseTags": null,
"Hostname": "",
"NotepadURLs": false,
"AdvertiseRoutes": [],
"NoSNAT": false,
"NetfilterMode": 2,
"Config": {
"Provider": "google",
"LoginName": "rafaelgarrido@gmail.com"
}
}
```

using tailscale ping works from the macos:
```
$ tailscale ping --verbose --until-direct=false -c 5 rafa-vps
2021/11/04 00:53:18 lookup "rafa-vps" => "100.111.81.66"
pong from rafa-vps (100.111.81.66) via 45.79.152.156:41641 in 12ms
pong from rafa-vps (100.111.81.66) via 45.79.152.156:41641 in 12ms
pong from rafa-vps (100.111.81.66) via 45.79.152.156:41641 in 10ms
pong from rafa-vps (100.111.81.66) via 45.79.152.156:41641 in 12ms
pong from rafa-vps (100.111.81.66) via 45.79.152.156:41641 in 11ms
```
logs:
```
default 00:53:42.986501-0400 launchservicesd CHECKIN:0x0-0xa80a8 8791 io.tailscale.ipn.macos
default 00:53:43.032237-0400 distnoted register name: com.apple.xctest.FakeForceTouchDevice object: io.tailscale.ipn.macos token: f4268 pid: 8791
default 00:53:43.077763-0400 IPNExtension ping(100.111.81.66): sending disco ping to [Q9inX] rafa-vps ...
default 00:53:44.094871-0400 IPNExtension ping(100.111.81.66): sending disco ping to [Q9inX] rafa-vps ...
default 00:53:45.111865-0400 IPNExtension ping(100.111.81.66): sending disco ping to [Q9inX] rafa-vps ...
default 00:53:46.128669-0400 IPNExtension ping(100.111.81.66): sending disco ping to [Q9inX] rafa-vps ...
default 00:53:47.144032-0400 IPNExtension ping(100.111.81.66): sending disco ping to [Q9inX] rafa-vps ...
default 00:53:56.463490-0400 IPNExtension magicsock: want call-me-maybe but endpoints stale; restunning
default 00:53:56.721716-0400 IPNExtension magicsock: STUN done; sending call-me-maybe to d:5141a06972a8c0c0 [Q9inX]
```

### Steps to reproduce

This is only happening with my macOs machines (tried with 2 different machines on my network)

I can ssh into rafa-vps.rafaelgarrido.gmail.com.beta.tailscale.net (100.111.81.66) from a Manjaro Linux laptop running tailscale 1.16.0

### Are there any recent changes that introduced the issue?

Nothing on my end... this problem happened with macOs version 1.16.1 and 1.16.2

### OS

macOS

### OS version

10.15.7

### Tailscale version

1.16.2

### Bug report

BUG-d0cdeac48b8a4900fb87e12b4977eee90f9bce5a6910845b3dbb3feb298b66d3-20211104044344Z-9c0349ae777479ac