Joe Sack answers a number of questions about intelligent query processing in SQL Server:
You have batch mode adaptive joins, but no row mode adaptive joins. Why?
Adaptive joins are more appropriate for scenarios where the join-input row count fluctuates significantly. Batch mode assumes a higher row flow vs. an OLTP low-row typical pattern. Row mode adaptive joins would likely be too prone to regressions. Batch mode on rowstore opens up adaptive joins for scenarios where we estimate higher row counts for join-inputs.
There are some good questions and answers in this set.