When designing a recipe, I often run into issues where all the options I've selected are valid, but because of something intrinsic to the database, like not having enough memory to execute the query with the indices as designed, the run fails on execution. However, it can take a while to catch these issues since I don't always want to run my recipes right away.
It'd be great if, without writing any data to the downstream dataset, the underlying query could be tested to see if it starts returning rows, and the results previewed, before actually writing any data to the output dataset. This would increase the speed of the feedback loop for recipe design and quickly inform me if I have issues. It'd also make validating the selected options in the recipe much easier.
The main error I'm interested in catching is the Teradata spool space error. The main validation I'd like to do is ensuring custom defined columns are working as expected. But in general, and for many databases, there are many other problems that could be caught by testing the query and previewing the results before running a recipe.