The serious part is: I have a little doubt on the viability of RIPEMD160, by the curse of "birthday paradox", approaching 2^80 computation doesn't seem to be too remote. It's just not economical (profitable). At present, the whole network can do 2^66 hashes per second, it takes 2^14 seconds - less than 1/4 day, to reach 2^80 computation. And remember, that is 2 times SHA256 on a block. ASIC's that are computing one-time random SHA256 should cost less and do more.
finding a hash collision for RIPEMD160 is impossible because the chances of it are astronomically small and it doesn't do you any good because you need to have a private key that produces a public key that is then hashed with SHA256 and then hashed with RIPEMD160 that produces the same hash! that is the combination of 3 impossible things!
2^80 is huge and what miners are doing with their 2^whatever hashes per second is not a hash collision find!
You don't need to go through the same steps of generating an address, since you know the public key space is 160 bit long, you just need to search 160 bit long private key space, in a hope to find a collision.
Example: If I know you are mapping a value in the space of one trillion, 2^40, to a space of 2^4 (0-F in hexdecimal), I just need to try all 2^4 numbers to see which one collides, I don't need to know or compute the original number in the range of 2^4.
Bitcoin is mapping 256 bit key space to 160 bit space, actually the public key space is 2^96 times smaller than its private key space, - much more shrinking of co-domain than the above example 2^40 to 2^4 (only 2^36 times smaller)
I agree 2^80 is a huge number, actually the world's fastest computer can do 200K trillion flops, that's close to 2^58, to generate 2^80 keys, that machine will need 2^22 seconds, which is 46 days. Then you will have about 50% that there exist collided keys inside this huge list of 2^80 keys. I see 2^80 is the magic number that will put Bitcoin on the brink of collapsing, sure we are not quite there yet, but are we too far from it? Considering the rate the computers/GPU's/ASIC's are produced, I don't think the scenario is too remote.