I'm not too concerned about how masking or hashing of demographic data in development skews data statistics and performance testing. When coding SQL or designing indexes, it's best not to make strong assumptions about the distribution of data, because that can drift over time in production anyhow. When writing new stored procedures, I'll typically unit test some of the more complex SQL queries in production and compare it's execution plan against development as a final reality check before committing the final version of the code.