cds.linear.com/docs/en/lt-journal/LTJournal-V24N4-01-df-SPICEDifferentiation-MikeEngelhardt.pdf
Funny, just last week or so I noted that gear integration produced a totally un-natural decay of an LC ringing oscillation. Mike says to always used modified trap.
Even modified trap needs a little help with high-Q circuits, specifically manually setting the time step.