Currently, HAXM supports max 16 vCPUs per VM.

Considering recent multi-core CPUs (e.g., Xeon scalable processor), I think the current limit is small. KVM supports max 288 vCPUs.

It seems on NetBSD, HAX_MAX_VMS * HAX_MAX_VCPUS pseudo devices are created at the module initialization, but on other platforms, a pseudo device is created when creating a vCPU (i.e., in the hax_vcpu_create_host()). So I think except for NetBSD, increasing HAX_MAX_VCPUS has no side effects.

QEMU uses the HAX_MAX_VCPUS. This is my patch for the QEMU to allocate the hax_vcpu_state as much as necessary (I tested this on the Linux host / Linux guest). I'm not sure other software using HAXM.
Is there anything to consider?