For passwords you're right, but at least it prevents theft of every password all at once.
For Bitcoin hardware wallets like Trezor, IIRC they either do or will support BIP-70 "Payment Protocol" payment requests that are signed with an X.509 cert, allowing you to verify the request on the dongle's screen.
Nothing stops a threat from just lying and waiting for you to expose a large number of passwords. Having one stolen doesn't raise red flags in itself.
I don't think signed addresses will be particularly effective. With the sort of key stores we have now, it seems pretty plausible that a bad actor to get a certificate that would pass on the Trezor device. It raises the barrier of entry a little though.
It should be just as effective as HTTPS is. Of course whether that's effective enough depends on who you ask. Yes, rogue/compromised CAs are occasionally a problem, but it generally works pretty well.
It doesn't hold up to SSL stripping very well. As we are working under the assumption of a compromised host, the absence of a signature on the Trezor when you don't expect one wouldn't raise any suspicion. The omnipotent host malware can remove all references to the payment request being signed from the payment gateway before the user sees it.
For Bitcoin hardware wallets like Trezor, IIRC they either do or will support BIP-70 "Payment Protocol" payment requests that are signed with an X.509 cert, allowing you to verify the request on the dongle's screen.