Credit card theft like you describe is already possible and already happens
The old method requires use of credit card blanks, a duplicating device, and the card itself doesn't look like the card when presented in person. In this case, as someone mentioned, Coin doesn't display the card #s on the front, so it's like a Card Not Present (CNP) transaction, but needs to be treated as if it was.
This doesn't bring anything new to the card skimming operation, but it simplifies, optimizes, and can in some ways facilitate it. They need some sort of ideally biometric authentication and/or a server that identifies when two Coin devices carry the same cards on it to avoid this sort of fraudulent use.
This device defeats all these human security measures by letting me use a startup bling device to try up to 8 cards at once without looking suspicious at all. "Declined? shit, my newfangled e-bling card must have messed up - try it again!" (while silently changing cards using the button). "No name on the card? Card isn't signed? You want my ID? But this is the hottest new toy! Promise it's fine!"
- Centralized location / evidence. A stash of blank cards and an ID printer have to go somewhere, and look suspicious to start with. The gear to skim cards and write them onto existing magstripe cards can be stored in a small space or on one's person and thrown away quickly; there's no centralized location. I believe hotel keycards are used because of their availiability. There was a sensationalized national news piece a few years ago about pimps giving women hotel keycards with credit cards written on them to buy gas with.
- Start-up cost. A couple thousand $$ in an ID printer is more than $0; that's what would divide street-level carders from professionals who probably wouldn't be passing overwritten cards at retail anyway.
It's not obvious how to prevent that without affecting the UX.