On Peripherals – Google is already addressing that problem with Cloud Print right now. The same concepts could easily be translated to other peripherals. Anyone whose tried to use it will tell you the concept is very rough right now but it has promise and several companies have agreed (in theory) to support it in their hardware with future versions.
On Large Files – I don’t know of anyone who doesn’t think our current bandwidth is just a fraction of what will be available to us 10 years from now. For video editors who need to use raw files I don’t think Chrome is appropriate but Google’s said as much by targeting the OS at the netbook/low-end notebook market to start. Again the point is to refine the OS through iteration (as Google does with all their products) so when editing video in the cloud is a possibility they'll already have the most refined OS for it.
Right now, these devices are not very cloud friendly. In general you plug your printer, scanner, etc. into a USB or Firewire port. From there the operating system needs to work with the printer through some kind of driver. With the sandboxing that I've seen with Chrome OS, it appears to be a rather daunting task to support a wide array of such devices.
Another issue is the bandwidth cost. Let's say I somehow make a video camera "web enabled". Instead of interfacing with the PC through USB, it connects through wifi or wired ethernet. From there you could upload your video to a cloud app that does video editing, or upload it straight to youtube. However, what happens though when it's an hour long video of a family wedding? Depending on the camera's storage format, this could turn out to be a giant gigabyte file, and will take a ridiculously long time to upload given standard broadband connections.
Until this issue has a good solution, I see it extremely difficult to achieve this "live in the cloud" philosophy. This is not to say it's impossible, just that it needs to be well thought out.