6to4 could've been released to billions of users, with no cooperation from the ISPs, just by home router manufacturers adding a few lines of code (maybe a few dozen, for health checking.)

Applications would still need to change, but apps change a lot faster than ISPs.

6to4 died because it was never the primary mode of addressing. Performance between 2002::/16 and the rest of IPv6 relied on anycast, which was unpredictable and therefore useless for real traffic.

Placing the 6to4 endpoint at the home router assumes that the home router has a public, routable IPv4 address, which doesn't really do anything to solve the problem of address exhaustion. At best it might have been a slightly superior form of NAT, but not enough of an improvement to offset the reliability and performance issues or the lack of economic incentive for ISPs (or anyone, really) to operate 6to4 anycast gateways and potentially invite non-customer, and thus non-paying, traffic to route through their network. Considered as a transition mechanism, the fact that 6to4 works better for two 6to4 LANs communicating over IPv4 (where packets can be routed directly) than between 6to4 and native IPv6 (which requires anycast proxies) was a real deal-breaker. It encourages a dependency on tunnelling via IPv4 rather than a transition to a native IPv6 internet.

6rd, fortunately, has none of these drawbacks, though it does require support from the ISPs.