I've uploaded the document CLB-018-1.doc to the Yahoo! MS/TP group.
This document addresses the inconsistency between the max-master and max-info-frames properties of the device object.
In my experience, it's more beneficial to MS/TP network performance that max-master be configurable somehow -- whether this be "via BACnet services" or some other mechanism. The BTL specified tests state that if the max-master property isn't writable via BACnet services, then it shall be fixed at 127. That works for testing purposes, but for devices in the wild, leaving max-master fixed at 127 is very harmful as this can cause undo Poll For Master delays. Unfortunately, the language in 135-2008 supports the view that it can be fixed at 127.
So, this is my attempt to clear this up and bring consistency to this language. I'm also happy to propose the language for the specified test document.