In the EM36ST i am thinking what is the actual role of those capacitors in the circuit ?

If the capacitor sees HF applied to it, then it is important to not exceed its ripple current ratings. Provided that PHE850 have about 10 times lower loss angle (tan d) than PME271, you have likely upgraded the ballast and won't have any issues

If the capacitor sees lamp ignition voltage peaks that reach its max ratings, then it may be more complicated because each capcitor type may react differently

Yes I'm not sure what the role of them actually is in this EM module, and why they didn't just use plastic film caps originally? I have another identical EM 36 ST that I changed the same caps in last year, and that is fitted in this fitting which seemed to work ok when tested.

The ballast is designed to last 10 years, and there is an underlying assumption (which is not always correct) that it is not stored for too long from manufacture to when it is put to use

The PME's will last 10 years, the problems with them start sometime at 20 years or older

Plastic film caps with the minimum voltage ratings (typical 275V) may fail already within the 10 years. The capacitor locations in the ST ballast don't suggest that they are EMI capacitors - They may be things like DC-link or even in the lamp circuit, although the type of capacitor does suggest EMI (all other capacitors in the ballast are not in X or Y positions, so typically manufacturers use there capacitors without X or Y ratings)

If then we ask why a bit higher voltage rated caps are not used, that's because the manufacturers don't deem it necessary to overdo if it seems to fit by ratings

(And hey, the electrolytic cap is typically rated for 15 years from the date of manufacture, as an additional limit besides the hours rating)

If they would want a ballast that is likely to last "unlimited" life, and assuming the manufacturer's engineers are aware of the PME failure mechanism, the capacitors would be indeed some plastic film but with 1.5x+ voltage ratings, which would also be bigger in size

In few cases a PME may also last for many years beyond the 20 year line, if it is working in conditions that prevent the failure from happening (see Perfektstart example). I don't think it was intentional (or that the failure mechanism was known at all) back when those Perfektstart ballasts were designed, but nowadays if a PME is used in a similar way, it may not be a significant limiting factor in the ballast's life. (HF ballasts are not enclosed enough, and dont have components that get hot enough for this to work for them)

And, you may be surprised how much common knowledge of us electronics amateurs (and those of us who became engineers) is completely alien to absolute most of engineers working in design of devices. This goes well beyond being familiar with failure modes of some components