Apply suggestions from code review

Co-authored-by: Daniil Gusev <daniil@quix.io>

Author: gwaramadze

docs/joins.md

@@ -283,7 +283,8 @@ The merging behavior is controlled by the `on_merge` parameter, which works the
 
 ### Limitations
 
-- Joining dataframes belonging to the same topics (aka "self-join") is not supported
-- The `backward_ms` must not be greater than the `grace_ms` to avoid losing data
+- Joining dataframes belonging to the same topic (aka "self-join") is not supported.
+- The `backward_ms` must not be greater than the `grace_ms` to avoid losing data.
 - Interval join does not preserve any headers. If you need headers from any side, consider adding them to the value.
-- Performance of the join depends on the density of the data. If streams on both sides move very fast (a lot of messages) then the performance may drop significantly.
+- The performance of the interval join depends on the density of the data.   
+If both streams have too many matching messages falling within the interval, the performance may drop significantly due to the large number of produced outputs.