Easy to reproduce with a simple pipeline:

node("windows") {
    retry(2) {
        bat("ping 127.0.0.1 -n 60 > nul")
    }
}

Manually abort the job during the first execution of the bat step - notice that the body of the retry is ran again rather than the entire job aborting.

It appears that the FlowInterruptedException caused by the CauseOfInterruption.UserInterruption is suppressed by the bat step which only kills the underlying process.