One more point I considered in my holistic analysis is that for most transactions we can't be anonymous. Thus anonymity is more suited to those who want to receive some payment anonymously and hide the funds there and extract them only to public funds in small morsels or to spend in other rare anonymous transactions (e.g. buying some gold bars from someone you trust won't reveal your identity).

In that case one might think you can just use Stealth Addresses (unlinkability) and run a full node to confirm receipt of funds anonymously. No need for Cryptonote, RingCT, nor ZeroCash. But the problem is the payer can be identified and be pressured to reveal your identity.

So this is why we need Zerocash to make the untraceability impervious to meta-data correlation.

But the problem with my proposal for ephemeral Zerocash mixers is that when we take the coins out of the mixer they can now be correlated to our meta-data (e.g. IP address, etc). So thus it seems to hide large funds and only take out small portions publicly as needed, will incur risk of losing those coins in my proposal, but at least they will be provably anonymous.

Anonymity is a clusterfuck. If we can't make trusted hardware, then anonymity is unprovable. Period.

So just give up on anonymity, or get busy trying to make hardware we can trust?

(or if Zerocash has developed a provably secure way to generate a master public key, which I doubt)