Well it does seem like an extremely high proportion of stale and duplicate work. I know this isn't related to hardware errors but there is one miner which seems to be producing a relatively high percentage of HW errors also. I didn't screenshot that as it's only about 1.6%.
You're misreading the graph. Hashrate and rejects have different scales, one on the left side, the other one on the right side. Under 1% of rejects is not high.