Re: Tested how many times?
Yes, you are supposed to use obfuscated data in test environments.
I would recommend that unless the system is entirely new production data is used where possible if you want things to actually work. I don't work for a bank, fortunately, but the systems I work on include :
Implementations where the code and data have changed over a decade, so what the data look like at year one are not the same as in year ten, for the same actions.
Data that have bypassed all validation, because validation wasn't included in a special out of bounds data submitter. The data were of course presumed correct when supplied.
Staff who have manually edited data, despite being told not to, because it's easier and quicker than doing it the right way.
Data that have been manually edited because otherwise it would need a convoluted code change, that has been tested and works for years, until a special new upgrade comes along that can't cope with the 'data in a state that is fine for everything except this' condition.
Code that has been incorrect for a short period, and tweaked data in ways that it shouldn't.
Tools that have the power to break things if they are used incorrectly, being used incorrectly.
Data being deleted when it is a dependency for something else
Data in the future or past because it was entered legally, but incorrectly.
Obfuscated test data would be unlikely to do any of the above, it's difficult to generate enough data to realistically test the system, and it's non trivial to anonymise all data as some have dependencies on others, and still possible to identify people.
For instance, if you tested your organisational chart modelling tool, with a certain Mr. Goreja Tray at the top, with 21 heads of department below him, another 294 'employees' in the organisation, supported by a third party organisation with 10 members, competing on 'deals' comprising 650 delegates including themselves.. I, for one, would have absolutely no idea where the source data came from.