A follow-up to this... we discovered that if we re-wrote the Service to have the start method block until the stop method is called, the Service is stopped cleanly on the machines that displayed the 1067 error code previously. I know the main documentation page says that if you use the jvm mode this is the proper way to code the Service to work, but the Service appeared to work fine without blocking on all the machines on which we did the early development on, so we figured we could get away without implementing the block. Still, would be curious to determine what's different on the machines where the shutdown failure occurs. We deploy with a common JRE instance so I would think this may be Windows configuration difference.
-- "There ain't much to being a ballplayer, if you're a ballplayer." -- Honus Wagner
