1. Double spends are devastating and easy, since they permanently split the tangle.
2. With no transaction limit, syncing from the beginning of time will take forever.
3. With no transaction limit, keeping up with network traffic will be impossible. (Especially on IoT devices.
4a. Nobody is going to use power and die space on IoT devices for the PoW chip.
4b. Or, alternately, if, as they claim, the PoW chip will take very little die space and very little power, the network will be destroyed outright by non-IoT PoW chips spamming the network.
5. There is currently a coordinator which confirms transactions. It is not P2P. If they remove the coordinator, I could write code that destroys the network by issuing TiB of transactions per day, making it impossible to sync/keep up.
6. Mesh networks of the type that they envisage deploying IOTA on are not widely deployed, and it's not clear that they will ever be widely deployed.
7. Tip selection does not converge.