Plan based on today's ACR call: A decoupled architecture involving communication by querying each other's metadata stores for now (later a message bus mechanism maybe introduced).

User(Curator) will use VA's UI to query sparql endpoint in Medici for metadata field "ProposedForPublication" to get collections marked for publication.
Rules to filter datasets to be moved from ACR to VA will need to be introduced later at ACR/VA.

Medici will query VA endpoint, based on collection's id (which was proposed for publication) in Medici (VA stores the Medici id - VA id mapping), to receive the DOI assigned for that collection and update the collection's DOI metadata field with this DOI in Medici. I will soon update information on the VA endpoint to be queried for DOI.

Update (based on oct 24th ACR call):

To get the DOI of a collection or dataset archived in VA from ACR, you can query the servlet at http://bluespruce.pti.indiana.edu:8181/dcs-nced/query by passing the id(encoded) in medici as follows (eg: ACRR Basic GIS coverages):

http://bluespruce.pti.indiana.edu:8181/dcs-nced/query/?q=resourceValue:(tag%5C%3Acet.ncsa.uiuc.edu%2C2008%5C%3A%2Fbean%2FCollection%2Fde62ed08-d61e-4cf2-ad22-41bf142e881e)

The returned result is an xml of metadata in the VA data model, describing the collection's metadata. You can then extract the value of DOI from the alternateId element:

<alternateIdentifier>
<typeId>doi</typeId>
<idValue>http://dx.doi.org/10.5072/FK2J105XM</idValue>
</alternateIdentifier>

I have added for 6 NCED collections(eg: ACRR Basic GIS coverages) in Medici, the value for field "Proposed for Publication".

These DOIs could be manually updated in the collections' metadata in Medici.

Or

A "DOI query" button could probably show up for those collections' Medici page which have "Proposed for Publication" field set. This button could then query in the above fashion , by appending the tag id as an argument.