A big problem of torrents (and IPFS, and most p2p technologies) is how to give proper incentive for people to contribute. There are "cheating" clients that don't upload back for example. And there are people that close the torrent as soon as it is downloaded. Well, you can sometimes rely on people's sense of belonging to a community; or you can sometimes shun noncooperating people from the community (as in private trackers).

I think dweb's main contribution is giving the option of monetary compensation for participating in the network. Perhaps such incentive wasn't needed when people were just seeding things they are passionate about, like movies and music, but it may enable more boring stuff to be decentralized as well. Hence, blockchain.

Most current p2p protocols are about sharing data in a decentralized way, but you can also execute algorithms on this data. That's what's being done with blockchain today (technically you wouldn't need a blockchain for this, but we already established we might want a way to give monetary incentive to peers, so, there's that)

This scheme may or may not work. I think it may have adverse effects, like when the web was taken down by advertising and commercial interests. But it also opens a path to a new kind of web.

About ddos: existing protocols have a big problem with scaling; an actual denial of service attack would make things much worse. That's why Lbry (which is a decentralized Youtube of sorts) is mostly accessed through a centralized site, Odysee - funnily enough, Odysee doesn't use any form of decentralization to serve the videos. It could make peers serve themselves using webtorrent or other protocol on top of webrtc, and it might some day just do it, but it doesn't at this time.