You can both have five minutes of work and some extra JVM speed juice over awk. The following Kotlin Script runs in ~2m46s on my M1 Mac. This is faster than the baseline and a bit more complex than it otherwise could be to avoid using too much RAM.

I think it's also more readable than the awk for people who don't already know the language whilst being of a similar length, but that might be just me. More usefully you can get type checking, autocompletion, use any libraries you want and so on.

    class Result(var lo: Double = Double.MAX_VALUE, var hi: Double = Double.MIN_VALUE, var total: Double = 0.0, var samples: Int = 0) {
        val mean get() = total / samples
    }
    
    val results = HashMap<String, Result>()
    Path("measurements.txt").useLines { it
        .map { it.substringBefore(";") to it.substringAfter(";").toDouble() }
        .forEach { (name, measurement) ->
            with(results.getOrPut(name) { Result() }) {
                lo = min(lo, measurement)
                hi = max(hi, measurement)
                total += measurement
                samples++
            }
        }
    }
    results.toSortedMap().mapValues { (_, m) -> "%.1f/%.1f/%.1f".format(m.lo, m.mean, m.hi) }.also(::println)

You can do this sort of thing with jbang, although for some reason Gunnar's baseline is slower. Maybe I'm doing something wrong.