When planning an IPv6 strategy is it possible to have link-local only addresses within the core transit network (sort of analogous to a BGP-free MPLS core)?
If not why not?
If yes are there any disadvantages to using this model; maybe the potential of making troubleshooting a little more difficult, such a traceroutes not working?
Are there similar deployment models that should be considered (such as a ULA core)?
It is certainly possible and has been done. Usually each device gets a global loopback address for example for iBGP sessions and remote management.
Take a look at RFC 7404, it describes exactly what you ask.
In theory is possible to use link-local IPv6 as gateway address, but IPv6 link-local accessible only from local subnet (it nonroutable), so then how to access equipment for management over IPv6 form other subnets?
