Does this effect show up only on Aluminium electrolytic capacitors, or also other types of electrolytics ?
All dielectrics.
The lowdown is here.
https://en.wikipedia.org/wiki/Dielectric_absorptionIf you do a search on "dielectric absorption" or " dielectric relaxation" or "soakage" or "battery action", you'll find plenty of articles on the subject going into finer detail than that wikipedia page. The reflow temperature effect is just a side issue of DA is all.
I think the reason why Dave didn't come across information on this DA phenomenon in capacitors was simply due to not knowing what the usual descriptive terms are for this related effect. I started by DDGing "zombie voltage in electrolytic capacitors" with, of course, no luck.
It took me quite a while before I finally came across those descriptive terms and matched them to my own experience when trying to select low leakage 1500uF caps to create an RC of 1000s of seconds using 10M resistors. I needed a matching response in a feedforward correction circuit to match the thermal inertia of an LPRO I was trying compensate for the parasitic heat flows that introduced unwanted variations of its internal thermal gradients despite all the insulation used to confine heat flow through only its base plate with a temperature stabilised fan cooled heat spreader to hold it to within 20mK of its set point temperature.
I gave up on using an actual RC for the job and spent the next week or so looking for "digitally synthesised RC delay" and variations thereof before finally coming across this ridiculously simple C++ one liner as described in this link
https://www.embeddedrelated.com/showarticle/779.phpIt seems my mistake had been in failing to use the search term "Single-Pole Low-Pass Filter"
Anyhow, the cost in processing cycles using an Arduino nano was very tiny with a time step of ten seconds and tau of 1000 seconds or more so it rather nicely did the trick sans the problem of DA effects in a feedforward correction cct where the tens of millivolts of DA would cause problems at the microvolt level on the ruby's EFC pin. Not a major issue in the 1200s RC LPF used by the PLL in my gpsdo as a feedBACK circuit element (2M with 500uF).
BTW, the reason why I know this is related to DA effect is on account the heat of my fingers was enough to vary the the "zombie voltage" I was monitoring with an SDM3065X on a collection of candidate caps I was testing for low leakage after leaving them shorted out for hours beforehand (hence my using the phrase "zombie voltage" because it was rising up from the dead).