If the value of the token ever gets far above the cost to mint, eventually more people will start creating it, and more will be produced, which, if demand stayed constant would result in the value diminishing. In practice, most tokens already do spend significant amounts of time paying out more than they cost to mine (which is why people mine them at all). Also, these effects take time to occur, so it is most likely that they would overshoot, crashing the value of the coin. Minting would then stop, but you already have a much greater supply than demand so there would be no incentive to mint more at all. You'd be stuck with the coins as worthless - you have no mechanism to make them increase in value.

Now, if instead of using some uninteresting hashing, you did genuinely useful work to create the coins, perhaps like gridcoin, then your system might make sense - since it's representative of an amount of computation performed.

Nobody in their right mind would hold it though - I presume you'd use it to pay for other computation, and there'd be a market of companies wanting to buy it in order to get their workloads computed.