Nothing near relevant now, but some of you young kids may see this eventually with time stretch algos. Note that time compression also suffers in a couple areas having to discard data, just not as glaring as when you try to stretch the wave file and end up with something full of holes. The excellent vid Maartian posted could have been served better by showing artifacts induced upon a sine wave, not a saw.
Like the birthing crisis of convolution reverb. Problems they faced, computation power, struggling technology. Predicted it would be decades before convolution could ever be used as a real time processor. Okay for post processing but not in the remote future for real time application. We done caught up faster than we thought.
What I think will happen and some of us may live to see.
Time stretch (resampling mode) has to cut out segments of the signal enough that the glitches of millisec silence become apparent with longer stretches, and a drag unless you are chasing that effect.
Future star wars models will do a look ahead thing based on the rate of dropouts, constant and predictable within the degree of the stretch. Resample the point of dropout and seam it smoothly and continuous over silence until the next wave segment kicks in. Way beyond our capability in the present, but will happen in time.
My opinion about the fate of noise gates which are still abysmal and barely functional.
These things will follow the FIR lineage. More accurately able to process a specific noise floor profile and extract it, without taking everything else out in the tonal spectrum as a casualty.
Bright future for our children's children's children. Maybe even before then.
John