This should be fixed. If you have jobs waiting, you can delete them and resubmit them and they should start running quickly.
Details for anyone interested:
2/3 of the cluster we were using for normal jobs went down last week and hasn't been restored yet. The remaining 1/3 has been operating at (or over) capacity since then. We've just added the ability to spill jobs over to another cluster when the first is full.
The first cluster still has a backlog of jobs in the queue that can't be automatically migrated to the second cluster. These will eventually run, or you can resubmit them and they'll automatically go wherever there's a free slot.