PMBus allows for things like different transaction speeds (100khz, 400khz), and other minor variations. We want this because there could be potential differences between the configuration (for Vendor A's part) and the interface (e.g. Vendor B's pmbus dongle) that could create conflicts that should be caught beforehand.
Specifying this should be optional, and an interface would generally default to the safest/slowest configuration unless it's safe to do otherwise.
This is different than, say, vendor-specific limitations that require a forced timedelay after issuing a certain command. That will be facilitated by a delay step.