Thanks for testing.
Did you copy and paste a large document? If so there would be a delay in rendering (which would be even longer on the demo which has debugging set to on as it prints the messages into console)
Based on my tests, around 77k words to copy and paste took about 2 minutes on a local instance of LT + n-grams. Most of that though was waiting for LT to get back to me.
On that front I consider 2 things:
a) Adding a progress bar
b) See if dividing it up when it is large and sending it in parallel helps on that front (though for private instances)
If not there is a 10 second delay between rendering on lines when using the public server. Private instances could lower that down to as low as 100ms, though probably best to set it to 1000ms (1 second).
I am not quite getting your 2nd statement. You mean like when you click below the red line it gives you error suggestions? The mouse should change to a question mark at eligible areas. Maybe a screenshot would help?