I figured someone, somewhere had already tried to write the code, and alas http://www.zynaptiq.com/unveil/overview/. 4096 samples of latency. Proposed use is in post, not live. Free demo. Target is early reflections only, not 'distinct echos'.
That's an interesting piece of software. It looks like it's targeted at film and recording production. I played with the sample a bit and while it works as advertised, it only allows you to hear and compare results in "real time". There is still enough processing latency to disqualify if from live use.
And as you twist the virtual knobs to adjust the amount of processing, you inevitably come to the point at which the processing "degrades" the signal to an unacceptable degree. So it's then "back of the knob and continue".