Yeah the problem is that the temp curve is two halves of an S with a long flat spot in the middle so there's no real way to estimate it. It does calculate a straight line best fit temperature ramp if you have at least 30 minutes of food probe data and the ramp is more than 1 degree/hr. The algorithm only considers the last hour of data to "ignore" that the graph should be more of a curve. This number is displayed next to each food probe's temperature in the webui. As Ralph said, if a high alarm is set, it will also calculate the time to get there based on the current linear increase. The number is sort of worthless though due to the actual temperature being non-linear.
Someone get to work on HeaterMeter neural net AI for temperature and stall prediction!