The draft is very well written, clear in the scope and the way to have a solution.
The draft is trying to address a base problem for network operator that is to understand where and why (root cause) a packet loss occurs in the network.
To solve this issue the draft is defining an information model to classify packet loss (both intended due to some policy and unintended e.g. due to congestion) and then a data model to implement the IM framework for network elementsand reporting the types of discards and where these discards occur.
General comment: as said the draft is well written but I appreciated the fact do not go directly with a data model but to separate the proposal for an IM framework and a specific data model for network element providing an higher level of abstraction and independence of specific implementation.  (e.g. YANG, gNMI, SNMPv3 etc.).
Section 2 provides terminology section. Comment: I would rather have a bulleted list for terminology than a simple format. 
Section 3 describes the problem statement,reporting what already provided by present models (RFC2863 and RFC8343)and what instead added in this draft that is the type of discards. The section is clear , no issue for me.
In Section 5 a YANG data model is provided. I did not review in details the YANG format (I'm delegating YangDoctor for that) but it would be good that I could find in the identities definitons of discard-class all the sub-types defined in section 4.2, reporting the types of packet discards in the IM. 
In conclusion I found the document very interesting and for sure very useful in an operator prospective as a way to take the adeguate actions to mitigate the impact of unintended packet loss.