I can confirm what using "export QEMU_CPU=max,pauth=off" on aarch64 is helping, the slowdowns are not present anymore.
For RISCV64 this is failing with riscv64-binfmt-P: unable to find CPU model 'max' If I use "export QEMU_CPU=rv64,pauth=off" This results in: "riscv64-binfmt-P: can't apply global rv64-riscv-cpu.pauth=off: Property 'rv64-riscv-cpu.pauth' not found"
