Oh yes, we did do waterfall wrong too. It's possible to be successful at either. But, as per my comment in the other thread, 'wrong' was beyond individual control and into the organisational or industry-normal. However - I do think the consequences of doing waterfall wrong were more expensive than doing agile wrong, especially in the short to medium term.

As you say, there is a middle position. Even if you think you're fully embracing agile, you have to pick a granularity. At some point requirements need to be defined, and at some point it needs to be tested. So that grain may still be a miniature waterfall of its own, which is fine. Start doing whole-feature kickoffs or release testing or tech debt mop-up phases and the grain automatically gets larger. We tend to be coy about how we describe this - scrumban? - but it's always a compromise.