The engineering challenges are hard, probably not impossible. There is already serious research in to this by Astroscale and others. What is impossible though, is to launch a constellation of things to deorbit other things without a ton of money. Someone has to be willing to provide that money.

If you want to remove a small piece of debris and you can build a fairly large sat specifically for that piece of debris, its not difficult but not crazy hard. Most of the required hardware is available.

What is really hard is making it cheap, light and versatile thing.

Yeah of course, de-orbiting or grave yarding a single piece of debris is doable for a few million dollars right now, but we are talking about a large fraction of 900K+ pieces. That's the really hard part.

Much like surface trash cleanup, the perfect is the enemy of the good. The goal is "less junk on the streets", not necessarily "absolutely squeaky-clean streets". First get rid of that rusty fridge that someone dumped in the woods before we deal with all the cigarette butts.

Let's focus on the more problematic, larger objects first, that are at risk of creating much more debris if hit.

What's the minimum size at which debris is dangerous?

Since everything is moving at orbital speeds, and generally at an angle to each other, i understood even small fragments move at a few km/sec relative to what they hit, so don't need to be particularly large.

True, but I imagine that, the smaller the object, the smaller the chance that it will be hit/it will hit something, and a smaller amount of fragments that the hit can generate.