Redshift column VARCHAR(MAX) question

A customer is looking at creating a redshift table with a large varchar(65535) column (although only a tiny percentage of data will be large), but our best practice suggests that use the smallest possible column size (due to potential memory waste).

They’d like to clarify:

1. Does that mean having a varchar column with large max size will only affect performance if we store a lot of large strings in the column?
2. Would there be any problems if say 90% of the data is only 20 bytes?

I found a support article mentions that trailing blanks can occupy the full length in memory during query processing, but still not clear enough to answer the above questions.