For one employer I wrote a ton of tools that used sqlite as an memory datastore for manipulating data. All of the tools were written in perl and I found that if the data had any complexity then you needed an expert perl coder to deal with the in-memory representation. However, if you simplified it down to a few SQL queries that updated/queried an sqlite database then even the most novice perl coder could use/maintain/enhance the tools. For really large datasets you could back the table with a file and get really good results - a lot of programmers can't deal with a dataset larger than core memory, sqlite handles it with ease.
I found something similar, we moved a lot of logic out of Java and into SQL and PL/PGSQL. We realised that most of our business logic is mostly data transformation anyway.
Suddenly, the more advanced non developers who were logging into github for other reasons, could understand the code, and even occasionally debug it or help us work through new features.
These people were domain specialists, so being able to tap into their skills and knowledge at the code level was amazing. And this was in addition to the very significant performance and code density benefits we achieved.
There were so many unexpected benefits from moving our logic into SQL that I would struggle to justify implementing future projects any other way.
