The main feature of dRAID is the integrated hot spare(s). What that means is the spare blocks are scattered through out all the disks, allowing the dRAID integrated hot spare(s) to act as part of the data+parity stripe until a failure occurs. That improves performance.
Technically, their can be more than one parity group in a single dRAID vDev... and the integrated hot spare(s) can service both parity groups. But it's complicated.
On loss of a disk, dRAID can recover very fast because the destination, (aka hot spare), is over every other disk. This eliminates the hot spare as a write bottle neck for restoration of health to the dRAID vDev.
I don't know if any performance details of dRAID without integrated hot spares.
And yes, RAID-Zx column expansion does not include dRAID.
For the average user, (aka not experimenting), I would think that dRAID should not be considered. Their are too many quirks, where standard ZFS RAID-Zx and Mirroring, (with various support vDevs like Metadata), are quite well known at present.
Here are some document links:
openzfs.github.io
dRAID vdevs resilver very quickly, using spare capacity rather than spare disks.
arstechnica.com