If I was given a /48 IPv6 address space, I would keep it as simple as possible. I’d separate my address space into equal sizes and not bother with various prefix lengths. There are just so many IPv6 addresses that creating /127 point to point subnets is just unnecessary.
Let’s say I was given 2001:ABCD::/48. I would take this and separate it into subnets like so:
That would give me about 65500 subnets with 2^64 hosts each. I would use the same prefix size for all of my subnets, whether they are end user subnets, point to point links, or management subnets. For most organizations this is more than enough address space to be able to do this.
Sometimes it’s difficult to disengage from the subnetting mentality of IPv4 whose goal was to conserve addresses. Here we don’t need to do that, and this is actually one of the things that simplifies IPv6 compared to IPv4, even though it may not seem simpler at first glance.
Now if 65500 subnets is not enough, and for some ISPs this may actually be the case, then you can always further subnet these subnets by using a larger prefix. For example: 2001:ABCD:0:1::/64 can be further subnettet using a /80 prefix like so:
This will give you about 65500 networks per /64 network (a total of over 4.2 billion) with 2^48 hosts per network. (BTW, that’s as many networks as there are addresses in the whole of the IPv4 address space!! It boggles the mind ).
So you see there is currently no need to employ an address conservation mentality, and you can keep things simple by maintaining a constant prefix across all your networks.
I hope this has been helpful!