Many new coins these days use seems to avoid PoW since they claim it's trivial to perform 51% attack by renting CPU, GPU or ASIC.
To avoid that problem, you can always track the strongest chain of Proof of Work headers in existence in SPV way. You don't have to follow transactions from other coins. But if you follow their hashrate, then you can determine the real Proof of Work in the world, for a given mining algorithm, and then adjust your network parameters accordingly.
Which means, that if you want to create for example some new altcoin, which could be mined with double SHA-256, then you have to receive 80-byte block headers from other double SHA-256 chains, just to calculate the difficulty in the right way. And the same is true for every other mining algorithm: if it exists, and is used anywhere else, then you cannot pretend that "whatever, blocks from other chains are invalid in my chain, so I won't count them, when I calculate my difficulty".
Bitcoin was created in the way it was, because there were no altcoins, which used Proof of Work before. Hashcash was used only as a spam protection, and it was based on SHA-1, and not double SHA-256. It used different algorithm, and there was no danger of "being reorged by hashcash miners". Which means, that the initial design didn't consider any other chains, so copy-pasting it 1:1 doesn't work, because if you start from scratch, you compete with existing huge networks (and Bitcoin had no such competition, so that design decision was good enough then, but shouldn't be repeated now).
But today, the situation is different. Today, if you want to make a successful Proof of Work altcoin, then you should count Proof of Work from other chains, because if you don't, then you will have the same problems, as for example NameCoin. If you have Merged Mining, then it can help a bit, but still, the design of your altcoin should guarantee, that you can reorg it, only if you can 51% attack the strongest hashrate with your algorithm, and not just reach 51% on your chain (also, that mistake is the very reason, why NameCoin was 51% attacked: that kind of attack, should be as hard, as doing it on Bitcoin, and to achieve that, it is needed to receive every block header, no matter what content is behind it, and count it to calculate "the real total difficulty", and not just "the local difficulty").
As Satoshi said:
I think it would be possible for BitDNS to be a completely separate network and separate block chain, yet share CPU power with Bitcoin. The only overlap is to make it so miners can search for proof-of-work for both networks simultaneously.
The networks wouldn't need any coordination. Miners would subscribe to both networks in parallel. They would scan SHA such that if they get a hit, they potentially solve both at once. A solution may be for just one of the networks if one network has a lower difficulty.
I think an external miner could call getwork on both programs and combine the work. Maybe call Bitcoin, get work from it, hand it to BitDNS getwork to combine into a combined work.
Instead of fragmentation, networks share and augment each other's total CPU power. This would solve the problem that if there are multiple networks, they are a danger to each other if the available CPU power gangs up on one. Instead, all networks in the world would share combined CPU power, increasing the total strength. It would make it easier for small networks to get started by tapping into a ready base of miners.
So, to achieve a successful Proof of Work altcoin, it is needed to fullfill that sentence: "Instead of fragmentation, networks share and augment each other's total CPU power". And the reason, why many altcoins failed, is that nobody designed it in such way. And as people saw, that for example NameCoin can be 51% attacked, they gave up the idea of Merged Mining, without realizing exactly, where it failed, how it failed, and what exactly NameCoin did wrong (they simply ignored Bitcoin headers, which didn't contain "NameCoin commitment", but they should instead blindly accept it, and use those headers to properly increase their difficulty, decrease their coinbase transaction, or perform similar actions, to measure "amount of coins per difficulty" properly).