I’ve only ever used Jool – because I’m vain and foolishly think I’ll one day have enough traffic to warrant multiple state-sharing translators – but I think Tayga has it right from their README:

Stateless versus Stateful NAT64

In situations where stateful NAT64 is required, TAYGA can be used in combination with a stateful IPv4 NAT such as the iptables MASQUERADE target. This allows the administrator a great deal more flexibility than if stateful NAT were implemented directly in TAYGA.

For known-sized subnets (eg security cameras), it can be perfectly appropriate to use stateless NAT64 so that old hardware can speak the modern Internet Protocol. But in other situations, stateless NAT64 is the only way to go and the Tayga developers knew to not reinvent the decades-old, proven stateful NAT44 mechanism in the Linux kernel.

And not could they hope to even compete with the kernel, as Tayga is based on the user space Tun/Tap driver. Jool, however, is in kernel, and can do some amazing (read: unusual) things. But just because Tayga only supports stateless operation doesn’t mean it should be ignored as a valid NAT64 solution, even if it’s not full featured.

IMO, the more ways we can get away from Legacy IP, the better.