I think it's important to separate the issue of what monads are/how they're used from the question when they should be used at all. While monads are very useful for working with various streams/sequences even in imperative languages, they are used in Haskell for what amounts for effects in pure-FP, and that use ("Primise" in the article) has a much better alternative in imperative languages. Arguably, it has a better alternative even in pure-FP languages (linear types).

Here's a recent talk I gave on the subject: https://youtu.be/r6P0_FDr53Q