There was a good bit of discussion about this particular issue, both here and elsewhere, back in 09. I'm not aware of any clear explanation.
Basics.....this from a very qualified individual who is not allowed to speak online:
.....the liquid pre-polymer molecules each contain 2 or 3 SH-groups (sulfur-hydrogen). SH groups in each molecule react in the presence of a peroxide (contains oxygen), split out water, and form an S-S bond (sulfur-Sulfur). This produces the chain extension or crosslinking. The curing agent (catalyst-accelerator) in the case of common aerospace sealants is manganese dioxide (MnO/2).
.....the following reference is suggested: Concise Polymeric Materials Encyclopedia, Joseph C. Salamone (ISBN 9780849322266)
A VAF contributor located a technical paper linking elastomer deterioration to peroxides. However, a careful read of the polysulfide sealant conclusions finds:
....From the available data it is uncertain whether catalyzel oxidation or acid catalyzed hydrolysis at the formal linkage is the primary degradation process. The latter mechanism is favored....
The manufacturer's rep discounted peroxides, but did report sensitivity to paint strippers.
Just of the heck of it I mixed sealant samples in correct and incorrect proportions. After cure, three samples were immersed in a 50/50 mix of MEK and lacquer thinnner. A fourth was immersed in ordinary drugstore 3% hydrogen peroxide. They've been there a long time now (since 11-16-09); none show any obvious change, certainly not reversion to goo. I am a chemistry idiot, so I'll leave conclusions, if any, to others. For sure 3% peroxide will reduce a new paint can to rust:
FWIW, there is no specification requiring the manufacturer to test sealant in avgas. The test fluid is called "JRF", for "jet reference fuel".
Nobody has ever offered evidence (a chemical process, physical mechanism, a measurement or test) to validate the "dripping avgas" theory. We can safely assume most of the avgas components evaporate, and thus concentrate the remaining component(s). That remaining component should be obvious if someone would simply hire a lab to run a softened sample through a GC/MS for comparison to a good sample.
Or you can debate a few more years
