The announcement is specifically about FIDO2 adding support for two additional things:
1) The ability to share FIDO credentials between multiple devices. Previously, it was implied and alluded to, but never stated outright, that credentials would be bound to an authenticator, like a MacBook's Secure Enclave, which FIDO calls a platform authentictor, or a Yubikey, which FIDO calls a roaming authenticator. Now there's explicit support for multi-device credentials. Apple recently added this feature in what it calls "Passkeys", a name that other vendors (but not FIDO) seem to be adopting too. This is net positive. Losing a device that was bound to a credential meant that the credential was lost forever. Now, as long as the credential resides in at least one device the user has access to, there's no recovery flow needed. Note that the vendor providing syncing services for these credentials does not have access to them. See https://support.apple.com/guide/security/secure-keychain-syn... for an example implementation
2) Expanded ability and commitment from vendors to use a roaming authenticator over Bluetooth Low Energy (this is already in the standard). And in particular, the ability to use a phone's platform authenticator as a roaming authenticator in a different device. This does not mean, as TFA implies, that you'll need a phone to sign in to services. Rather, it means that for services that allow or require FIDO credentials to sign in, a phone is now an additional option to present those credentials. You can still use a Yubikey, TouchID or any other way you interact with your existing TPM.
I understand that people are concerned about new authentication standards backed by big corporations who have a history of locking users out of their platforms and services, but the current state of secure login is dire. FIDO2 is an incredibly well designed set of protocols to prevent phishing, credential reuse, and several common causes for account compromise. It was clearly designed with that in mind, at the expense of usability. These are notable and incremental improvements to enhance the usability of a standard that is head and shoulders better than existing alternatives like passwords, but still has some ways to go in terms of functionality. Personally, I'm very excited about FIDO and WebAuthn, and some of the improvements I'd like to see in the coming months are:
a) The ability to share passkeys across vendors, including the ability to implement a "sync fabric" as some folks in the WebAuthn working group have called it, so it's interoperable beyond the major vendors. b) For these vendors to strengthen their own log in experience. Apple only allows their own TOTP implementation and SMS fallback to authenticate to iCloud. I'd like to use WebAuthn exclusively here, so I could back up access to my now-precious Keychain that holds all my FIDO credentials with a YubiKey. c) A better story about backing up security keys. Implementing a) would give us that. Devices that can be initialized with a given seed like some common hardware crypto wallets would give us that, albeit not without introducing changes to the threat model -- you have to store the seed and input it somehow -- and https://www.yubico.com/blog/yubico-proposes-webauthn-protoco... would give us that as well. d) A better story for usernameless. The current methodology to have a user initiate a usernameless login and picking the right credential is a UX mess, and I don't believe I have actually seen it implemented in a production site. I'd love to be shown an example!