Starting from the assumption that you are already dealing with PDFs as the incoming format, PDF/A has a number of advantages, especially in terms of encapsulation. A properly created PDF/A will ensure that there are no dependencies on external resources by embedding them inside the file. This is particularly important for fonts, but also applies to other media and content. Similarly, a well made PDF/A should also meet a certain standard in terms of embedded metadata.
PDF/A-3 represents a kind of middle-ground hack here, by allowing arbitrary resources to be packaged alongside the document. This NDSA report
gives a excellent overview of the issues. My personal opinion is that, at present, the PDF/A-3 cure is worse than the obsolescence disease, and I would prefer a flavour of PDF that enforces embedded fonts etc. but discards nothing.
When the incoming format is not already PDF the situation is similar to the above, but much more extreme. It's even harder to know whether what you are throwing away is important, and so keeping the original becomes even more necessary. For example, conversion to PDF/A does not necessarily imply that the external dependent resources of the original item will be picked up correctly. Finally, a decision to normalise multiple formats to PDF implies a lot of assumptions about modes of future discovery and use of the material, and I'm not convinced those assumptions will stand the test of time.