Sometimes, but not always. We deleted the ones that were not, and finished the ones that were.

As far as noticing these by hand: my current project is about 600,000 lines of hand-written code, not including several libraries we also author and use in it, nor other projects I'm also heavily involved in. I haven't even looked at all the code yet, much less do so on any regular cadence. Forgetting is something you plan on and address up-front, not something you just let happen.

Why make lists that will never be completed?

Or put another way, there is no pain that can't be avoided by ignoring the backlog.

You can only ignore these TODOs until your build fails, then you have to reflect on whether to do them, push them into the future, or delete them.

I think this works because it really makes backlog grooming a group activity where everyone is involved, so everyone feels they have a say in setting priority.