Such is partially the case with propagation of conventional path DML's through MV's even within the same database. Insert 100,000 rows in a single statement and you get 100,000 single-row inserts against the MV log. The logged change vectors are propagated more efficiently though, not through one DML per changed row.
Direct path insert is more efficient and avoids this issue, but I haven't used it in the context of replicating between databases.
Specifically, a large DML operation in the primary database is converted into individual row-level DML statements in the logical standby. Our testing indicates that any DML that impacts over 100,000 rows in a single transaction will slow down linearly (or is that exponentially?) in the logical standby. For example, an update to a column in all of the rows in a million-row table in the primary database will take about 3 minutes but, in the logical standby, it will take 5 hours for the resulting million update statements to complete.