Does that mean that 3% of seizures require intervention?
I should have put over 97%. But yeah for a generalized tonic-clonic seizure lasting longer than 5 minutes is called "status epilepticus" and that is a risk of lasting injury and thus warrant intervention. Just not the kind that bystanders are able to provide.
I would imagine the probabilities aren’t independent, but if they were, the probability of someone staying in the 97% for 10 seizures in a row is 73.7%. 20 seizures in a row drops the probability to 54%.
Yup, probabilities are not independent and if those clusters of seizures happen minutes to hours from each other, the risk of injury increases further. Also, with more and longer seizures the epilepsy tends to become increasingly harder to manage. But epilepsy comes in many shapes and forms so it depends on the specific kind. The adults who die of epilepsy usually don't die "unexpectedly" meaning they have certain comorbidities that increase the risk of dying (e.g. heart disease), etc.
What are you basing this statement on? (the causality part)
Of course. Occam's razor still applies here. But knowing how resistant ME/MDs (depending on state) to put "epilepsy" as the cause of death, the "it was seizures" explanation remains unusual. BTW, I ended up looking into it, so the chief ME's report is still pending. So it's not like anybody has final conclusions and everybody's speculating here.