Weird situation going:
2x RB3011 connected through a stretched VLAN.
ether1 goes to the WAN and ether10 is connected to the datacenter.
I have a VRRP interface (called vrrp-v6) running on both RB's, on ether10.

Just like with V4, I configured each RB with 2 IPv6 addresses: one for a "host" IP address and the other for the VRRP address.

But when I try to ping the master, I can't ping the vrrp and the host address.
The only way for me to be able to ping the master, is to remove the host IPv6 address. Then the VRRP address is pingable.

The backup unit (who has the VRRP IP disabled) is pingable through the host address.

When I re-add the host address to the master, both vrrp and host IP are not reachable.
Though there are reachable in the same /64 prefix. Outside that /64 prefix (which part of a bigger /48 prefix), I can't reach the vrrp and host IP.

When I look in the routing table when I have the host IP address on it, it says the /48 prefix in the datacenter is reachable through interface ether10...and when I remove the host address, that route is reachable through interface vrrp-v6 (which is bound on ether10).

VRRP interace on both RB's: R1 (backup): R2 (master): Could someone help me solve this mystery? I want to be able to have AND the VRRP IP AND a host address on both RB's.

This sounds like a transit network with nothing but routers on it....
You should strongly consider switching to OSPF instead of using first-hop-redundancy protocols. (just my $0.02's worth there)

Also, I haven't messed with vrrpv6 yet (but I'm going to try it out in a lab after posting this) but I'd think that you might get better mileage using link-local addresses.
Hosts / client routers should set their default GW to be the link-local of the vrrp-v6 interface, and the internet routers use the link-local address of the (I'm guessing vrrp interface) downstream gateway.

You'll notice that OSPFv3, SLAAC, RIP, etc all tend to use link-local as the gateway for everything.

Your assumption is correct, it is indeed a transit network with 3 routers (2 RB3011's and 1 FortiGate).
I know OSPF or BGP would be "better" but this installation, the routing back to the RB3011's has to be static and to one IP only.
That's why my choice was VRRP-v6.

You mean I have to route the 2001:xxxx:xxxx::/48 to the link-local address of the Fortigate? Or to the link-local address of the RB interface.
If the latter, then what interface should I choose? The vrrp interface or the physical (ether10) interface?

And the weird thing is...I recreated this in GNS3 with CHR and the exact situation works as I wanted (vrrp + host address).
Same firmware level (CHR en RB3011), so I don't get why this doesn't work on the RB.

I have solved the problem, thanks to this topic.
After changing the VRRP IP to a /128 (and not advertising it) on the master, the host address + the vrrp ip became reachable.