Policy-based routing not using correct gateway/not populating gateway following gateway flap

crgbt

Some background info:

pfSense verison: 22.05-RELEASE
Hardware: Netgate SG-3100
Connections:

Starlink on mvneta2
VDSL via A&A (aa.net.uk) over PPPoE (via mvneta0)
L2TP tunnel via A&A (aa.net.uk) over Starlink (via mvneta2)

Output of system aliases in /tmp/rules.debug:

#System aliases

loopback = "{ lo0 }"
STARLINK = "{ mvneta2 }"
LAN = "{ mvneta1 }"
AAISP = "{ pppoe0 }"
GUEST = "{ mvneta1.102 }"
IOT = "{ mvneta1.103 }"
AAISP_STARLINK = "{ l2tp1 }"
AAISP_MODEM = "{ mvneta0 }"
Tailscale = "{ Tailscale }"

Output of gateways in /tmp/rules.debug when everything is working:

# Gateways
GWSTARLINK_DHCP = " route-to ( mvneta2 100.64.0.1 ) "
GWAAISP_PPPOE = " route-to ( pppoe0 81.187.81.187 ) "
GWAAISP_STARLINK_L2TP = " route-to ( l2tp1 81.187.81.187 ) "
GWSTARLINK_AAISP_FAILOVER = "  route-to { ( mvneta2 100.64.0.1 )  }  "
GWAAISP_STARLINK_FAILOVER = "  route-to { ( pppoe0 81.187.81.187 )  }  "
GWL2TP_STARLINK_AAISP_FAILOVER = "  route-to { ( l2tp1 81.187.81.187 )  }  "
GWAAISP_L2TP_STARLINK_FAILOVER = "  route-to { ( pppoe0 81.187.81.187 )  }  "

The relevant user alias from /tmp/rules.debug:

table <AAISP_Outbound> {   172.25.1.221  172.25.1.222  172.25.1.223  172.25.1.224  172.25.1.225 }
AAISP_Outbound = "<AAISP_Outbound>"

(That gateway IP is not sensitive; it's the PPP endpoint which is published on the A&A website: https://support.aa.net.uk/Server_List)

PBR is set to route a set of IP internal IPv4 addresses always out of GWAAISP_PPPOE with the following rule:

pass  in  quick  on $LAN  $GWAAISP_PPPOE inet from $AAISP_Outbound to any ridentifier 1653729485 keep state label "USER_RULE" label "id:1653729485" label "gw:AAISP_PPPOE"

As both my VDSL and L2TP connections are provided by the same ISP, the gateway address is identical (81.187.81.187), but this routes out the right interface after a fresh pfSense reboot.

The issue is that after a random number of gateway flaps on AAISP (pppoe0), pfSense loses the gateway for GWAAISP_PPPOE, even though it shows as up in the UI and the metrics for ping responses are updating. This could be after one gateway flap, it could be after ten - there doesn't seem to be a common scenario, but eventually pfSense does lose the gateway. When it loses the gateway, the output of gateways in /tmp/rules.debug look like this:

# Gateways
GWSTARLINK_DHCP = " route-to ( mvneta2 100.64.0.1 ) "
GWAAISP_PPPOE = "  "
GWAAISP_STARLINK_L2TP = " route-to ( l2tp1 81.187.81.187 ) "
GWSTARLINK_AAISP_FAILOVER = "  route-to { ( mvneta2 100.64.0.1 )  }  "
GWAAISP_STARLINK_FAILOVER = "  route-to { ( mvneta2 100.64.0.1 )  }  "
GWL2TP_STARLINK_AAISP_FAILOVER = "  route-to { ( l2tp1 81.187.81.187 )  }  "
GWAAISP_L2TP_STARLINK_FAILOVER = "  route-to { ( l2tp1 81.187.81.187 )  }  "

As you can see, pfSense doesn't re-populate the GWAAISP_PPPOE gateway after the connection comes back up. I've tried restarting dpinger and manually reconnecting the PPPoE interface (via Status > Interfaces), and neither work; the only thing that starts traffic routing back out of GWAAISP_PPPOE is to reboot pfSense entirely.

I'm kind of at a loss; everything seems to be recovering correctly, it's just pfSense seems to not pick up the gateway after a random number of gateway flaps.

I appreciate any pointers you folks can give, as I've tried everything I can think of.

Thank you in advance!