Underlying the temperature effects on lifetime and failure is ---- the removal of HEAT from INSIDE the capacitor.
The capacitor has only 2 leads. Ensure those 2 leads go directly to WIDE pieces of metal foil. Since usually one of the leads goes to GROUND, you have an excellent opportunity to perform thermal engineering and keep that capacitor much cooler.
Also ensure the ungrounded lead has wide foil, and that foil is immediately passing OVER GROUND, for about a centimeter. A centimeter is a useful distance (in FR-4) to dump heat into an underlying plane. You could include an otherwise useless 1cm by 1cm bulge on the HOT LEAD, right by the lead, to dump heat into the underlying PLANE (Gnd or otherwise).
You might strongly focus on "thermal reliefs" around the lead, if the cap is throughhole mounting. Thermal Reliefs will DOUBLE THE THERMAL RESISTANCE of that region of foil.
Copper foil of standard thickness is 70 ° C per watt per SQUARE OF FOIL. Thus a trace 10mm long and 1mm wide has 700 ° C of thermal resistance.
Its your task to explore, on a quadrille pad, assuming each little square is 70 ohms of resistance, with one amp needing to enter the grid, and realize your PCB layout has the lifetime of that capacitor in your hands.
If you have moving air, then the CASE can dump heat into the air. Or --- even directly radiate to the other oomponents and chassis.
I recall old TVs with 2" by 4" aluminum electrolyitcs and 2 or 3 or 4 mounting tabs that twisted to lock, and then might be soldered. Good heat flow. But we don't have that anymore.
What do do? With vertical mount of 2_wire caps? Perhaps BEND over the wires, to provide a large wide heat flow path.
This might work for NASA one-of-a-kind power supplies, but not for automated assembly;
On other hand, automated assembly likely has only some warrantee 5_year expectation, not 100 years in orbit or out past the solar flux limit past Pluto or surviving another 2 minutes in a Venus probe-to-800 ° C satellite.