You lose bandwidth by going to a lower servo PID rate. If that doesn't
matter because the system you're controlling is slow, there is no price
in going for a lower control speed.
With a sampling rate of 250Hz, you can actually measure up to 125Hz. In reality
the limit is usually between 1/5 and 1/3 that frequency before difference between
a sampled and a time continuous system become apparent.
So with 250Hz you're approximately limited to 50Hz of control
bandwidth. To me that seems terribly low. On my own lathe I have a
resonance at approximately 300Hz. Since I have the servo thread running
at 4500Hz (because that's the maximum rate the encoders I use support)
I was able to eliminate that resonance.
On the other hand, what do you lose when going for a higher sampling
rate? The only thing I can come up with is processor time, but that
seems hardly an issue on a modern PC. And I'm doing this on an RT_PREEMPT
kernel, so it's not even Xenomai or RTAI.