50 usec sampling -- why not?

I can only get my slink to echo kenwood KRX1000 commands accurately by using 50 usec sampling. It works great at this setting.

So what is the down side? (I'm assuming that there is a downside, since the default sampling rate is NOT 50 usec) Is there insufficient processor resource to sample at 50 usec and support the full complement of IO? Does it help if I disable some of the inputs?

I'm thinking that if the higher sampling rates do cause problems if lots of inputs are enabled, my software (using the C++ class library) can dynamically disable inputs under certain circumstances.

Regards,

Dave Thomas