Problem stsetch SNF 2010-08-17 10:25:45: eating resist, not silicon!

We are releasing cantilevers with through-wafer etches from the backside stopping on buried oxide. Our wafers are 450 um thick and 7 um of standard resist typically works fine. We have released 3 wafers in the last 2 weeks and the resist etch rate has increased progressively while the Silicon etch rate has decreased... Here is our data:
8/4 - 7 um resist, through wafer in 4:02, resist survived fine
8/9 - 7 um resist, through wafer in 4:22, resist lost at ~4 hrs on part of wafer
8/16 - 10 um resist, resist totally gone by 3:30, wafer not near done after 4:15 etching.