I agree:
Our current objective is clear: compute something useful->get proportional BOINC credits->get proportional GRC rewards.
I also agree that:
"Compute whatever->get arbitrary RAC->get arbitrary mag" dilutes that objective.
The WU hosted on GRCStarter@home could be anything. It depends on how its operation is structured:
Perhaps one set of WU is crunched at a time. At any time, someone can propose a different set of WU for GRCStarter and the community votes on whether or not to switch to that work.
Perhaps there is a ledger of WU ready to be processed on GRCStarter. The work is added to this ledger by some process -- voting in, proposal, whatever. There is a poll that runs for a block of time... days, weeks, months, years, whatever. People can change their vote to support different WU sets while the poll is running. The amount of time allotted to specific WU sets is proportional, or some relation, to that sets support in the poll. When the poll ends, there is some process that determines the next ledger of WUs -- maybe it's the sets with the highest support from the previous poll + sets with the highest support in a second poll regarding proposed sets. A poll is then created for the new ledger.
Perhaps GRCStarter is only for researchers which put together proposals for getting their work crunched on GRCStarter and the community votes on the best proposal.
Perhaps there is a limit to the number of work units GRCStarter can run for a project and a time limit each project can run its WU on GRCStarter -- once x number of cobblestone is produced or x number of hours is met, or whatever, the project can no longer host WU on GRCStarter. This means that if a project wants all the attention from new users and philanthropic magsters it must put the effort in to convince the GRC community it is a worthwhile project.
Perhaps, one day, GRCStarter gets integrated into the Gridcoin blockchain and stakers choose which WU get processed through blockchain consensus.
These processes removes the risk of someone buying out, or monopolizing, BOINCs main resource: its processing. This risk is present in a rain system, but can be managed through regulation.