Implementing DQS/MDS in an Existing DW Environment

I'm looking for guidance on implementing DQS/MDS into an existing DW environment.  Specifically:

1)  When implementing, do you use it to correct dimensional data already in the DW to improve spelling issues; capitalization inconsistencies; format inconsistencies (date format, state abbreviations, etc.)?

2)  If you do correct the old data, do you do it as a Type 0 update?

3)  If you don't correct the old data, is it safe to say that the benefits of implementing DQS/MDS will be recognized from the date of implementation moving forward?

4)  If you already have data cleansing and matching being performed as part of the ETL process, does it make sense to replace it with DQS/MDS?

5)  What recommendations and/or guidelines do you recommend to follow? 

Appreciate any feedback.

Thanks you

Bigguy