There is another effect in PWW: The wire is drawn to size and annealed as a -final- step in the mill, and if you look at it even under high-mag SEM photo, you'll see that the wire surface is just about theoretically, perfectly smooth - there is no edge discontinuity like you will always have on a flat, rolled out foil with rough edges. Again: foil resistor can't really be final-annealed the same way once the foil is bonded to a substrate, and the shape - by definition - will always have relatively very rough edges compared to a smoothly drawn wire.
Noise is always generated at surface discontinuity, or anywhere there is internal stress and strains, and any place in a conductor where there is a -change- in conductivity. A round wire doesn't really have edge discontinuities or sharp corners that give that conductivity change - and it's a shape that can be made from very high-purity and highly consistent alloys. Just don't kink the wire or leads!
Once you roll out foil and attach to ceramic or glass, those electron pathways are constantly changing - that isn't a design that lends itself to zero stress or virtual zero-stress -changes-. Some of the film resistors do have noise that's pretty low, but never really achieved non-existent 1/f noise as in PWW.
You can even try binding foil to a ceramic, and then polishing the foil surface to less than 1u rms roughness, and it'll still have more 1/f noise than PWW. Beckman tried this in the late 60's and 70's, and basically couldn't compete with PWW's at DC. They could get the noise lowered over the standard foil rolling, but the danged things cost about 20x more than PWW's, and was a failure.
PWW have always been, and are still really the only practical resistor device design that has -zero- 1/f noise down to true DC.
Remember - this has been a problem with foil resistors that's been known since high-performance instrument amplifiers were needed - even TI was trying to make film resistors in the early 60's and could never reach true zero 1/f noise like PWW give you. Not a lot has changed today; since the same physics still apply.
NOW: On a new PWW bobbin freshly wound, yes there is a period where you'll see internal winding stresses relaxing and getting into final position - but that is generally a change in resistance or apparent TCR - that isn't true 1/f noise. Like all precision components there is an initial break-in time where the devices have to self-relieve assembly stresses and find their final stable running operating point and equilibrium. PWW as well as Bulk Film resistor can benefit from an initial thermal cycling to help gain stability - the change in temperature and generally higher temperature help equalize the strain throughout the material. No different than thermal-cycling precision mechanical components, castings, bearings etc.
You'll also see this in a new 732b: When those are new they have to be re calibrated much more often until the Zener and it's film resistors settle in. Compare that to something like and older 732a with PWW's inside - generally less noise & more stable to boot now that these are decades old (IF they were well taken care of). Fluke had a real hassle getting those 732b's to behave over time...that laser-trimmed resistor network isn't all rainbows and butterflies (over time) compared to the PWW's.
Both of those models really show how hard it is to make a stable 10V from ~7V though. There's not an easy answer to that one, no matter what kind of resistor you put in an Op-Amp feedback path.
By the time you build a DIY 10V Vref device that's truly as stable as a 732, you might as well have bought a 732...