setTimeZone("Asia/Karachi")

// control
dlsSwitchOver = createDateTime(2022,6,1,0,0,0)
writeOutput(dlsSwitchOver.dateTimeFormat("yyyy-mm-dd HH:nn:ss")) // Both CF and Lucee: 2022-06-01 00:00:00

writeOutput("<hr>")

// experiment
try {
    dlsSwitchOver = createDateTime(2008,6,1,0,0,0)
    writeOutput(dlsSwitchOver.dateTimeFormat("yyyy-mm-dd HH:nn:ss")) // Lucee: 2008-06-01 01:00:00
} catch (any e) {
    writeOutput(e.message) // CF: Date value passed to date function createDateTime is unspecified or invalid.
}

Note the behavioural difference between CF2021 and Lucee there. Lucee goes "yeah, OK, you say

2008-06-01 00:00:00

, but on that day the clocks go forward right then, so it's actually

2008-06-01 01:00:00

". CF goes "yeah, OK, you say

2008-06-01 00:00:00

, but on that day the clocks go forward right then, so I'm gonna error".

Both seem reasonable in a way, but the behavioural difference is not good. Initially I side with CF from a perspective of "garbage in / garbage out", but I checked what Java does in an analogous situation, and Java does what Lucee does. This example is runnable on CF (but not trycf.com unfortunately, because reasons):

testZonedDateTime = java {

    import java.time.format.DateTimeFormatter;
    import java.time.ZonedDateTime;
    import java.time.ZoneId;

    public class TestZonedDateTime {

        public String getControlDateTimeAsString() { // 2022-06-01 00:00:00
            return getFormatedZonedDateTime(2022);
        }

        public String getDaylightSavingSwitchoverDateTimeAsString() { // 2008-06-01 00:00:00
            return getFormatedZonedDateTime(2008);
        }

        private String getFormatedZonedDateTime(int year) {
            DateTimeFormatter formatter = DateTimeFormatter.ofPattern("uuuu-MM-dd HH:mm:ss");

            ZonedDateTime dlsChangeOverTime = ZonedDateTime.of(year, 6, 1, 0, 0, 0, 0, ZoneId.of("Asia/Karachi"));

            return formatter.format(dlsChangeOverTime);
        }
    }
}

writeOutput(testZonedDateTime.getControlDateTimeAsString() & "<br>") // 2022-06-01 00:00:00 NB: hour is **0**
writeOutput(testZonedDateTime.getDaylightSavingSwitchoverDateTimeAsString() & "<br>") // 2008-06-01 01:00:00 NB: hour is **1**

I also tested PHP, and it behaves the same as Java and Lucee. I got a mate to test C++ and it also is clever enough to work out what the code is asking for. My conclusion is that this is a bug in CF. Someone on that Stack Overflow raised a ticket for it (https://tracker.adobe.com/#/view/CF-4213572), but the underlying cause is slightly obscured by their example. Initially I piped-up saying "not a bug", but... I've changed my mind now. I'm left wondering if there is any legit basis for the current CF behaviour to not be considered a bug?