Well "A computer" could be any TCP/IP enabled device nowadays, including a router. So no, it's not reasonable to look at traffic coming from an IP, and say that traffic originated from the physical location of that IP, just because the LAN beyond that router is opaque.
I mean, several ISP IP addresses are involved with routing, but "everybody knows" they're shared machines. But again, simply because a trace route stops at a certain machine doesn't mean the routing actually terminates there - for most cases this is not true. It's just that the person behind the router with NAT often owns all devices on the LAN/opaque side. Using this correlation alone is dangerously inaccurate.
A search warrant leading to a malware infected machine is acceptable in my eyes. Again, tying traffic to a location is fine, as long as the correlation ends at the computer that originally appears to be the source of the traffic (even if it ultimately isn't) is all that is targeted until evidence can determine the individual behind it.
Search warrant for what? All TCP/IP enabled devices? Or the router? An IP address doesn't even uniquely identify a single device.
(Also given how search warrants are executed in the US and how long they can seize these devices for, I don't know if I'd say it's "acceptable". I'm in favour of setting the bar high to protect the innocent.)
You can most certainly narrow down an IP address to a particular ISP customer. Is it possible that they have an open wifi? Yes. Is it possible to narrow it down to a single member of the household? Depends! Is it possible that a computer at the destination is being used a proxy by the real attacker? Yes! But it's certainly not the blackbox that the EFF is trying to portray it as.
It's totally appropriate to execute a search warrant based on IP logs. A search warrant doesn't mean that any particular person is guilty, just that there is probable cause that there is information about a crime at a certain location.