From what I've read we don't have confirmation that the ratio between the work units is expected to change past today.

Uplinger, I see your reply - your link posts to what you've written about work unit credit and the differences in job size, but it's not explicit why this would result in boundless CPU tasks but less work for GPUs than anticipated.

The pipeline consists of many parts but this is the high level flow.
a. Researchers identify ligands and targets they wish to run
b. Researchers create batches for us in the form of 10,000 ligands against a given target (the target doesn't change much)
c. WCG download these batches
d. WCG builds and loads batches by combining ligands into the same work unit that gets fed to boinc
e. THE VOLUNTEERS DO THE HEAVY LIFTING :) :) :) :)
f. Results are returned to WCG for validation
g. WCG assimilates the results and creates a nice package for the researchers, this can only be done when all results in a batch are complete
h. WCG send result package back to researchers
i. Researchers do post analysis on the results.

In this pipeline there are many parts to consider... storage, bandwidth, cpu, and researchers time. We try to make sure that all of the stages are running as efficiently as possible as to not to cause a backlog at any stage.