I'll elaborate on @pericynthion's answer a bit, with regard to what the "noise" may be, and speculate on the cause.
Despite the visual appearance in the question, a plot of apogee is not any smoother than one of perigee. Currently you show one plot with a range of 24 kilometers and the other with a range of 1,000 kilometers. Below, I've plotted the incremental changes from one TLE to the next of the perigee, apogee and average of the two as a proxy for semi-major axis (e.g.
apo[1:] - apo[:-1] etc. in Python). It's a little bit like the derivative except I haven't normalized to the uneven time increment from one TLE to the next.
You can see that the semi-major axis more stable than either apo or peri, indicating that much of the "noise" is in the determination of the eccentricity. The third plot is the change in eccentricity.
It's difficult to imagine a physical process that can "tickle" the eccentricity in this particular way, making it sometimes less circular and then more circular immediately afterward. If it were space-plane shaped, perhaps there's a way, but these objects are not space-planes.
Instead, what I believe is happening is that the reentering spacecraft is being observed mostly from a single location on Earth. If you look at the argument of periapsis, it's extremely stable as it should be. So this one location on Earth, wherever it may be, might always be seeing a similar section of the elliptical orbit. Mean motion can be extracted with high precision and independence from the other parameters from timing, but without several samples around the ellipse, the other parameters can have a high degree of correlation. Correlated parameters in fitting limited data can create all kinds of errors and noise.
Incidentally you can see that the change eccentricity of PSLV-C39 has a negative offset, as well as the plot of its perigee but not apogee. This is characteristic of what happens to an eccentric re-entry. At first the drag impulse lowers apogee until the orbit almost circularizes. Then the constant drag lowers the orbit much more quickly, as described in this answer and illustrated in the final figure.
below: Simple simulation of a spacecraft in elliptical orbit with low perigee. The orbit first circularizes, then decays. See this answer for a more thorough discussion.