Inconsistent warning of mismatched redunancy classes

[valarauca]

System information

Type	Version/Name
Distribution Name	Debian GNU/Linux
Distribution Version	12 (bookworm)
Kernel Version	6.8.12-8
Architecture	AMD64/x86_64
OpenZFS Version	2.2.7

Builds are distributed via ProxMox v8.3 so the -pve trail on zfs versions & kernel version were ommitted.

Describe the problem you're observing

The error

mismatched replication level: both disk and draid vdevs are present

Is highly inconsistent

For example creating a pool like

zpool create -n -m none -t pool -oashift=12 \
    log nmve4n1 \
    draid2:8d:11c:1s sd[a-k]

invalid vdev specification
use '-f' to override the following errors
mismatched replication level: both disks and draid vdevs present

we pull the old switcheroo on the order:

zpool create -n -m none -t pool -oashift=12 \
    draid2:8d:11c:1s sd[a-k] \
    log nmve4n1 
    

would create 'draid2:8d:11c:1s' with the following layout:

    draid2:8d11c1s
      sda
      sdb
        ...
      sdk
    logs
      nvme4n1

Describe how to reproduce the problem

1. log before draid -> fails
2. draid before log -> works

One more thing

If a metadata special device is used (for example mirroring drives).

Ordering also matters.

zpool create -n -m none -t pool -oashift=12 \
    draid2:8d:11c:1s sd[a-k] \
    log nmve4n1 \
    special mirror nvme1n1 nvme2n1 \
    ; echo -e "\n$?\n"

invalid vdev specification
use '-f' to override the following errors
mismatched replication level: both disks and mirror vdevs present

1

Now I get a totally different error

zpool create -n -m none -t pool -oashift=12 \
    special mirror nvme1n1 nvme2n1 \
    draid2:8d:11c:1s sd[a-k] \
    log nmve4n1 \
    ; echo -e "\n$?\n"

invalid vdev specification
use '-f' to override the following errors:
mismatched replication level: draid and mirror vdevs with different redundancy, 2 vs. 1 (2-way) are present

1