Cyclone III Active Serial Problem

I've never seen that kind of problem before... How is your clock signal? Are there any significant overshoot/undershoot? The Cyclone III is rather picky with overshoots when using 3.3V I/O, but still it shouldn't cause that kind of defect. Are you sure that it really takes the 3.3V down and that it isn't a measurement problem (induced by overshoots on 3.3V clock that would end up on the supply line?).