Because, i have a requirement, where my query result to grid, needs to run for more than 5K limit, and i have found out an alternative by running it in, batch mode.
The 5000 row limit on the Query to Grid is due to memory. Anything stored in a Grid Variable is landing in memory. If you have a job that tries to land more data into a Grid/memory than what the instance has it will crash. 5000 rows doesn't seem like much but if you had 5000 times 20 jobs that run at the same time, you have the same memory issue. It's really all about putting safe thresholds around memory usage on the Matillion instance so that it's unlikely jobs crash the instance. I hope this helps.
What I found works best is chunking up the data into arrays (or another data type depending on use case) and processing the chunks at a time using iterators.