Thanks for the replies. After spending an evening reading, figuring things out, I think that this function (speed comping) although a good idea, for my workflow it's actually causing me to spend more time rather than less to edit clips.
Actually I rarely do 'true' comping, i.e., clip a note/word from different phrases/lines. And when I do this, it's at most between two or three clips. If I really need to patch a bunch of notes/words to comp a phrase, it never really work out for me (sounding natural and fluid), for this situation I would re-track that line/phrase.
What I do most of the time is selecting between a bunch of short phrases (clips), phrases with some clear break point like a line of a verse or a guitar runs. I also have bunch of alternative takes to see if something fits better. For this, this implementation of comping doesn't fit. It would auto split all my alternative takes and I have to 'heal' each clips. Once a clip is trimmed, why not just discard the end bits, I have to lasso the bits and delete them manually as they cluttered up the view.
Maybe for future release this can be considered as 2 types of comping, first level is for selecting clip among many and a second level for going in to slice and dice (what it's designed to do now).