[sgarland]

> What's a real world situation where completely the wrong type could be written to a column?

* App with incorrect schema definition is deployed

* App written in TS or Python has type checks disabled

* App written in JS does math, and its fun parsing system decides that 1 + “1” == “11”

I’ve seen all of these.

> You can run a simple script to clean up the "horrific" data.

Depends on scale, and the tables / columns. If there are billions of rows, and the columns with incorrect data aren’t indexed… that’s a bad time.