So it turns out there is 2 Virtual IP bugs with if_pppoe. One was fixed and patched I mentioned earlier. The other one with in my opinion is nasty, is not fixed, and there wont even respond on their redmine, I have actually noticed lately they are not responding to a lot of redmine issues now, they seem to be only picking up on specific things.
So the unresolved bug is if you have more than one IP, and add the IP as a IP alias (it might also happen on CARP), then if the PPPoE interface needs to be taken down, it actually stays in an UP state, and you just end up with no WAN connectivity and the device stuck until there is manual intervention or a reboot.
Since this is on the actual kernel device, and the source code has not been released, then people like me have no chance of resolving it ourselves, and I expect because me and one other person using CE have reported the issue, no one on plus, and not a mass amount of users, they are just going to leave this broken code as the norm moving forward.
So as an example a ifconfig command to shutdown the interface simply doesnt work, it will take it offline, as it removes the primary WAN IP and any routing config, but the interface itself will stay there in a UP state with any IP alias's still attached.
The real world issue here is if there is some kind of outage on the internet service, then there will be no automatic PPP recovery. It will need manual intervention, either to reboot to flush out the dormant PPP interface, or disabling reenabling of WAN in the GUI which also seems to properly clear it out to allow a new session to start.
So in the end it was something that affects people who use IP aliases. Pretty wild that VIP's were not tested at all for this.
The old PPP does this via MPD software, so a lot of it was handled external to the kernel, its now almost all kernel. This is also why the new PPP is far less configurable as well.
