My current environment is a medium-sized database for a lightly-customized commercial application. We use full production data for development and testing.

In my days at an ISV, I dealt with obfuscation and pared-down datasets. We tried to replicate customer environments as closely as possible, but this can be tricky when your customers have deeper pockets than you. Sometimes the most important part of a rollout was to remind the install team to verify their customers had current backups.

I'll echo many opinions here. The key is to run final tests in an environment as close to production as possible.