It's hard to say definitively if functional bugs may be introduced, or if UI testing is enough, without knowing a lot about the application.
For example, a text resource that is displayed in the UI, may also be passed along to another application that isn't prepared for a translated string. This would usually be bad development practice, but it happens.
As Joe gets at, there is a potential for localized strings to cause new functional bugs. In my opinion, the risk is low, probably very low.
I'll go ahead and stick my neck out on this one and I'll say that the primary goal of my own localisation testing is to uncover interface errors, not functional errors.
I think that the extent of your localisation testing should be directly tied to your schedule and where it ranks with the rest of your priorities and the associated risks.
IMO, if you're using pseudo-translation or localisation then you should be able to feel in extreme strings which will maximize your opportunity for error. As far as digging for new errors in a translated system, I would be careful not to let this become your primary objective.
9 out of 10 people I prove wrong agree that I'm right. The other person is my wife.