[mediocregopher]

A key component of the hash algorithm used by blockchain is that difficulty can be adjusted, such that the algorithm ensures that the difficulty to mine the next block increases/decreases based on how long it took previous blocks to mine. In this way the network can attempt to maintain a predictable rate of blocks being mined as mining power fluctuates.

So for this to be a good computation for a blockchain to use it would have to have a way to modify difficulty in a predictable way.

[MathYouF]

This is only a problem if you care about predictable rate of blocks being mined right?

What reasons besides predictable inflation are there for desiring that property?

[positr0n]

It also has to make the networking part of Bitcoin fit within physics. From eyeballing the difficulty graph it looks like if you fixed the network difficulty in January 2010 then a new block would be mined every few picoseconds at the peak hash rate this year.

[mediocregopher]

In the case of a fork in the chain the network decides the winner based on which fork ends up being longer. If blocks are being mined too quickly then forks become more likely and the state of the chain is more volatile as the current "winner" keeps flipping around.

Blocks being mined too slowly is a more obvious problem: less transaction throughput.