Yes and no. Assuming the board is alive in the first place (which is a very big IF), I take 10MHz variation from session to session as acceptable. Which can also be down to variation in NB cooling (due to insulation) or weather on Jupiter. I've noticed that some of these boards are extremely PSU-sensitive. Also, if you shut the PSU off, then on first cold boot (or run up until the first major crash) you tend to get slightly better clocks.
From where I am sitting, I am actually baffled as to how people get these boards to do 600+ 32M with Wolfdale on non-LN2 cooling. It was never possible here in a reliable way, even with CPU under single stage, so I even stopped trying and am using a Conroe to pretest boards on single stage. Yet, the same Wolfdale setups can easily do 640-650 5:8 with CPU on LN2 all day long.