TRACE_EVENT_ABI

I would have thought that many uses of tracepoints (like hacking up a script to trace down an immediate issue) would not really need a stable ABI, beyond perhaps one or two kernel versions. Other applications might like a stable ABI, but not necessarily a "forever stable" one (an ABI of the sort "we won't break it if we can reasonably avoid doing so"). Which applications would really need a longterm stable ABI then, and what sort of tracepoints might they need?