A cellar-c1 node crashed in a way which caused a very important load on all other nodes; this is causing general slowness and an elevated error rate.
It should go back to normal gradually and will not take more than an hour at the most.
EDIT 17:00 UTC: Error rate and performance is back to normal