Although documentation states that Varchar are supported for Incremental reflection refresh , however on using the varchars, even for incremental refresh Dremio scans through complete dataset (i.e. reads all the records) instead of only identifying new records. This mean this is almost as good as full refresh since all the records are parsed through.
in the first screen shot below - when I do a reflection for first time 277,128 records are scanned (I/P)
https://docs.dremio.com/acceleration/updating-reflections.html
in the second one(triggere refresh for incremental reflection) below we can see that it took the same time as above and scanned the same number of records although the output is zero records, since no new records were added.
Ideally the incremental refresh should have been almost instantaneous if there were no new records.
Are there joins in the query that you are trying to refresh? Can you please send me the profile of the reflection refresh that uses full refresh?