Fluid dynamicist here. While I have not read this article in full, it's worth noting that there is no reason to believe the critical Reynolds number for pipe flow is a constant. There are well known experiments which have produced laminar pipe flow at a Reynolds number of about 100,000, 50 times higher than what this piece claims is possible. This is done through careful vibration isolation among other techniques, to eliminate as many potential disturbances in the flow as possible. This idea is very old. Reynolds himself was able to produce laminar flow at a Reynolds number of about 13,000. See here for a review: https://www.annualreviews.org/doi/10.1146/annurev-fluid-1221...
The suggestion that these "puffs" are produced and die at the same rate a particular Renolds number needs to assume a certain production rate, which is going to vary from setup to setup. Consequently there is no constant transition Reynolds number for all setups. I think the birth and death rate idea is okay, but I suspect the popular level writer oversimplified the result. I have not read the original paper but recall a talk that I think was by Hof on the same subject, and I recall that they are aware of this.
(Also, if it was unclear, this critical Reynolds number only applies to long pipes. Any other changes can affect this greatly. You can have laminar flows at Reynolds numbers of 10^5 or higher in some applications. I see this mistake more often than I would like...)
My PhD is in transitional flows, in particular the transition from laminar to turbulent boundary layers under elevated freestream turbulence — so flows like that over the roof of your car, but you also see them on gas turbine blades as well as subsonic aircraft wings.
We typically refer to these puffs as turbulent spots. They manifest in many ways. Lots of researchers generate them artificially with a wall disturbance however they can occur all by themselves. A boundary layer velocity profile can become unstable vis T-S waves [0] and via a complex mechanism generate turbulent spots. They grow spatially until the boundary layer is saturated. On the other hand freestream turbulence can buffer the outer part of the boundary layer resulting in a streaky flow near the wall. These streaks can interact locally and generate spots of their own bypassing the T-S process entirely. However after the onset the growth of spots to saturate the boundary layer is remarkably similar. You can download an enormous DNS database of just a single flow case here if you like. [1]
The suggestion that these "puffs" are produced and die at the same rate a particular Renolds number needs to assume a certain production rate, which is going to vary from setup to setup. Consequently there is no constant transition Reynolds number for all setups. I think the birth and death rate idea is okay, but I suspect the popular level writer oversimplified the result. I have not read the original paper but recall a talk that I think was by Hof on the same subject, and I recall that they are aware of this.
(Also, if it was unclear, this critical Reynolds number only applies to long pipes. Any other changes can affect this greatly. You can have laminar flows at Reynolds numbers of 10^5 or higher in some applications. I see this mistake more often than I would like...)