Pure cpu calculations are fast, for fpu side we have a soft fpu emulation so it’s slower than it could be.
In ppc qemu emulation hardfloats aren’t still implemented, anyway it’s quite fast for majority of jobs, especially with modern x64 cpus.
The real slowdown is in graphics rendering, it really causes a bad user experience.

Qemu was born with flexibility, accuracy and portability in mind, speed isn’t it’s main goal.
Consider that most part of users use it in kvm mode where the cpu emulation part (TCG engine) is completely bypassed.
One of main goal of qemu is build VM on big servers for very big IT companies.
It’s developed mainly by these companies and most part of features are implemented for their primary purposes.

Anyway I consider qemu the best option to preserve AmigaNG experience in a next future, for sure this already happens for old Mac ppc machines.

Qemu needs to be improved in amiga boards emulation accuracy and speed related to gfx and fpu parts.
In the past mos team patched it to run morphos in a better way but these patches was never shared with community.
Probably for commercial reasons due mos license, maybe for other reasons too, I can understand so no problem at all.