RIFA PME capacitors are not notorious for the fact of them failing. They are "notorious" for the fact of making it visible

The most common other type of capacitor used nowadays are polyester X2 capacitors (from multiple manufacturers). Many of them last only a handful of years before failing, except they fail silently by electrically burning away the metallization on the film

Daim and Mex are some brands of plastic caps that tend to fail rather quickly. I've seen many of them which have lost their capacitance within 10 years of use. There are more. The only good thing about those, is that they don't go up in flames.

As you say, it is of the utmost importance for the engineer to check the datasheet for what applications they are suitable. Many of these are used as a series capacitive dropper. Slowly but surely they then lose capacitance and eventually cause the device to quietly fail. Philips had a major recall for their Senseo coffee machines that failed in exactly this way.
There are other capacitor types with different dielectric which are more resistant against big voltage spikes you can find on the mains. I don't know which ones. FKP, MKP, MKT...

Either way, for anyone replacing these things in lighting fixtures, get an X2 or Y2 type, and preferably a Wima, Panasonic or other good brand. I wouldn't put another Rifa PME back into it. It might last another 20-30 years which for some may be plenty, but if that fixture is sold to another person who doesn't know it, they might simply throw it away if it starts smoking.

My parents have thrown away a BEAUTIFUL kitchen mixer in my childhood because its suppression cap burnt out. They didn't know how to fix stuff and that it's not the motor that burnt out, so even if you get it to a professional repair person it would be a cheap fix, and i was 10 or so so i could take it apart but i didn't know what caused the smoke. It was a very heavy AEG machine, early 70s, very powerful. I somehow remember the chassis as being made from ceramic, but it may also have been white bakelite.

Medved :

But this problem will also exist if the power is interrupted 2 times, the first time extinguishing all HID lamps, the 2nd time stopping the motors with "unloaded" capacitors of the entire lighting installation...

Cold HID lamps have low arc voltage (can get as low as 20V for the first few seconds in an 100V-arc HPS). At this time the ballast pulls higher than normal current, at very low inductive power factor. A higher value capacitor will reduce the current draw in this stage

But there are some reasons why NOT to overcorrect the power factor this way :

1.
The purpose of power factor correction is to prevent reactive current going in the supply lines, causing unnecessary losses and loading of the power line

The problem is not only for the lighting circuit itself, but for the main feeding cables for electrical installations that power also other loads - which are already "undersized" due to diversity factor of those other loads. (The diversity factor of lighting is 1, but of entire installations can be anything like 0.5..0.7)

So it makes the most sense to optimize the capacitor value to correct the power factor as precisely as possible in the long run when the lamps are warmed up, not in the short period when they are cold

2.
The circuit conductors and circuit breakers can take some overloading for short periods, which is adequate for letting HID lamps warm up (even if it starts at 2x current, as long as it goes down fast in the first few seconds after that, and stabilizes at 1x within a few minutes)

3.
The capacitors have extremely high inrush current at the moment of connecting to the line (the first AC half cycle), which is proportional to the capacitance and circuit impedances

Higher capacitance makes the circuit require bigger switching gear (switches, contactors) to withstand this inrush, and may get close to the limits of magnetic tripping (short circuit tripping) of circuit breakers immediately at switch on, especially if "B" breakers are used (as is common in some places in Europe)



The capacitors in European gear are normally sized to correct the power factor to about 0.9L to 0.95L (L=inductive) for the condition with warmed up lamps at full power, and rounded to the nearest standard capacitor value

This in theory leaves some more capacity to go before getting to PF 1.00 or even more before 0.9C (which may be still acceptable), but probably not that much, and i wonder where the oversized capacitors actually are on this scale

It is generic problem for all paper capacitors. It is practically impossible to really prevent any moisture to seep in it, unless it e some special hermetic construction (molded plastic/epoxy cases aren't). With older vacuum tube electronic (old radio, TV, audio gear) a leaky capacitor can easily destroy the tubes, so it is often way better to just replace at least the critical ones (interstage coupling from anode to the grid of the next tube) before even attempting to power the thing up, if those capacitors are some of the paper type (paper wax, these Rifa ones,...).

I think in Europe you are not allowed to oversize the power factor correction.
One of the problem is a safety hazard: If capacitive net load is present on a network and there are some induction motors running a lot of inertia, a power cut upstream will make the motors turn into generators, the capacitive net load will keep them excited and create very large overvoltage till the motors spool down.
That is the reason the PFC's are slightly lower than required, if there is a need for some tolerance. Undercompensating makes the motor lose excitation, so the voltage collapses quickly, so no overvoltage.

RIFA PME capacitors are not notorious for the fact of them failing. They are "notorious" for the fact of making it visible

The most common other type of capacitor used nowadays are polyester X2 capacitors (from multiple manufacturers). Many of them last only a handful of years before failing, except they fail silently by electrically burning away the metallization on the film

Many of the polyester (and even the better polypropylene ones) have an explicit warning in the datasheet not to use in capacitor/zener divider applications. There is just too much "activity" inside, right from the start, for them to work reliably for any length of time in an application that expects a somewhat stable current source. They get used as EMI filtering capacitors because nobody (on the manufacturing side of home and office equipment) really cares about EMI after the device passed the initial testing

By the time a PME capacitor fails -  by its external potting cracking from age of the material and initiating the failure mechanism, it is mostly still ok electrically

(polyester Y capacitors must be more reliable than their X counterparts against failing shorted, so there maybe indeed they last - I have not checked the Y ones that much)

Polyester capacitors sometimes die in a non passive way, when their self healing mechanism fails, mostly when constantly a bit high line voltage meets a bit low quality capacitors. They keep melting and smoking and extruding melted plastic from holes burned in their case for a long while

All i have ever seen from a PME is cracking and a puff of smoke for a couple seconds, even when the remains of the capacitor remained connected to voltage immediately after the event



What i do with old equipment with PME capacitors in it - Acknowledge that for equipment from the 80s or 90s they lasted 30 to 40 years, and replace them with brand new identical PME for the next 30 years

The RIFA PME capacitors are still made nowadays, though there appear to have been some minor changes in the capacitor appearance, probably from when they moved the factory from Sweden to Finland. In particular the appearance of the metal foil cover over the actual capacitor roll, which is the 1st layer preventing the external cracking on the surface from propagating deeper inside. Maybe they actually resolved the problem with the potting material, but we will know only in another 10..20 years

The ultimate choice of capacitors for lasting forever is polyester (and other plastic films), but with big voltage margins - Like using 440V capacitors (instead of the usual 275V) for 230V applications. Maybe they do it in some very high end industrial equipment, but i don't recall anything like that in the average consumer or "light industrial" stuff



Most lighting ballasts have rated lifetime of 10 years. Indeed, the good ones can last many times that, but this does mean that the manufacturer won't be putting conscious effort into making them last longer than that

RIFA PME capacitors have been used through the 80s in Magnetic PerfektStart ballasts made by Knobel Ennenda (and Eltam). There they are hidden inside the relay case, wired between the input 230V phase and Neutral after the lamp filament

I have found several of them in the late 00's, dating from around 1981, so were 25 years old at the time. The capacitor inside is discolored and cracked from being for years right next to a hot heating element, but i never seen them short out (which would blow the lamp filament, so be noticable from outside by looking at installs with those ballasts, which existed here well into the 2010's). I think what prevented them from failing there despite being cracked, is that they were not exposed to the external environment, but to the clean and dry enclosed space inside the relay

For home use, if you do not need a formal sticker, that is necessary if you need legal backing of your measurement accuracy (but on the other hand that could be a requirement of the insurer if it involves instruments used to ensure workplace safety, like your handbag multimeter you are using to check whether a power is present before working on something electrical)...
I think you may do calibration check yourself with stuff you (will) have (a generator, oscilloscope, multimeter).
Normally with modern instruments this usually ends up just as a confirmation they are still working correctly, unless it is really some specialty equipment, drifting away out of spec practically never happens.
The only exception here is, if something gets broken in them, but then we are talking about fixing it first and all becomes way more complex...

To get real precision ethalons, some gadgets may help:
A GPS modules use to output a precission clock signal (10MHz) when locked onto satellites. Most common GPS chipsets do, many modules sold on web do use these and have that signal available. With them you have an atomic clock precision frequency standard at home, for few bucks. Just need to set it up so it can receive the signals.

A generic quartz analog clock movement generates precise 2s pulses and can be checked (you can observe how much that clock drift over time and take that as a correction)

A common HCxxx logic generating square wave signal is good for generating exact rectangular shape pulses to check the frequency response flatness of the oscilloscope input channel.

For voltage and current you need some external standard, if you want to ensure the whole specified tolerance of your multimeter. But routine check by comparing multiple meters you have is good enough to flag the malfunction, very rarely a defect turns into just small drift, often it gets way off...

As opposed to standard capacitors, they have paper dielectrics which break down over time. Always have done, always will do, regardless of the way they seal them. With one exception: paper submerged in oil, in a soldered shut metal can, typically actually lasts pretty long.
Any of the tar, wax or epoxy covered paper capacitors *will* fail. This is why old tube radios almost always need recapping.

Ceramic or plastic-dielectric caps don't do this.

Rifas are good at their job for as long as they aren't leaky. They perform better than some styles of plastic capacitors while actually being a fairly cheap option. But the failure mode of plastic capacitors is much safer than the failure mode of paper capacitors. Plastic and ceramic caps just don't burst into flames at the end of their useful life.

But imho there is no excuse to still use paper capacitors in the modern world. Even in the 2000s, those Rifas made in the 1980s were already burning out. The engineers should have known and not used the damn rifas. Even if they don't cause a fire, especially in big lighting installations they can cause a building evacutation which can cause significant loss of productivity that day (so, "lose" money).

These RIFA capacitors contain an epoxy casing which cracks over time allowing moisture to enter into the capacitor, make contact with the paper dielectric and boom.

How these RIFA capacitors blows up? These aren't electrolytic capacitors.