Derek Bickerton and John McWhorter both have studied pidgins and creoles. A general observation is that communities that have to deal with other communities who don't speak their language, but still have to communicate resort to using pidgins, creoles which are simplified, grammatically speaking. McWhorter says English lost it's complex grammar when Vikings arrived and couldn't deal with the complexity.
On the other hand, island communities that don't have to deal with other people can tolerate arbitrarily complex grammar since children have a spectacular skill for acquiring their first language as compared to adults. So as an isolated language accumulates difficulties, there is no pressure to make them go away, they just get added to the pile of obligatory complexities.
Another feature of Island speakers in specific is they are more likely to use a "towards the shore/away from the shore" as a grammatical marker, ref "When "North" Isn't Actually North: Geocentric Direction Systems"
EDIT:
Large phoneme inventory. Chinese, a lingua franca has a small inventory and a small inventory of syllables. Rotokas has a small phonetic inventory. Someone would have to graph this to see if there is a weak relationship.
Inflections (really, any morphology), irregular forms (lexicalizations), culture-specific complexities-- I'd predict more of all of these for an isolated language. Proving it to satisfy skeptics would take a few academic papers. English's "2nd cousin twice removed" system, is complicated, but English is a lingua franca. Hard to say beyond generalities.