OK, I was wrong. I had a look at several weeks of pressure in "detailed" reports. Looking at it day by day, I found that there are periods of these sawtooth pressure jumps. Some days a pick or two. Other days, I found periods of an hour or so with as many as 6 "teeth" and than smooth for next several hours.
I looked for the corresponding period in the flow reports and than it hit me - the pressure graph in the flow report is charted in the scale of 5 - 25 which is compressed but, the timeline is really stretched. In effect, in there, I never noticed any sawtooth pattern. On close examination, these sawtooth are from a very dull saw! The picks are not picks - they flatline for anywhere between 20 seconds to a minute with gradual raise and fall.
Now, I'm thinking that something must be triggering these raises. Not just machine blindly checking now and than if there is an improvement at higher pressure. If that was the case, I'd expect these checks to be present at regular intervals and they are definitely not. They seem to be preceded by changes in the breathing pattern / flow waveform. OTOH, not all changes in the pattern are followed by a pressure pick... Oh well, luckily people who designed the algorithm for these machines do know more about it than I do.
Everything I post on this board is nothing more than an opinion expressed by an apneak. Normally, it's based on facts and experience but sometimes, I may get things wrong or not have all the facts.
I reserve the right to change my mind. Why? Because tomorrow I may know better.