There are many common instances where ETT is indentifiable.

The three most common ones are:

1. Whenever ATE (Average Treatment Effect) is identifiable through adjustment for a set S of covariates.

2. Whenever treatment is binary and ATE is identifiable by any method whatsoever

3. In linear systems, whenever the total effect is identifiable.

The proof takes two lines, and also appears in chapter 11 of causality 2nd edition.

You are right, identifiability here means without restrictions other than those encoded in the DAG.I have not read the papers by Hernan and Robins in Epidemiology 2006; and Robins, Vander Weele and Richardson. Perhaps you can give us a pointer and a summary.

]]>