Incorrect JobDataMap when scheduling multiple jobs at once - quartz-scheduler

Question

Given a job definition (excuse my Kotlin) that stores data from trigger JobDataMap into job's JobDataMap
#DisallowConcurrentExecution
#PersistJobDataAfterExecution
class ProductUpdateEtlJob : RunEtlJob() {
companion object {
const val PRODUCT_IDS = "productIds"
}
override fun execute(context: JobExecutionContext) {
val additionalIds =
context
.trigger
.jobDataMap[PRODUCT_IDS]
if (additionalIds != null) {
context.jobDetail.jobDataMap.merge(
PRODUCT_IDS,
additionalIds,
) { old, new -> (old as List<*>) + (new as List<*>) }
} else {
super.execute(context)
context.jobDetail.jobDataMap.remove(PRODUCT_IDS)
}
}
}
Superclass details are not relevant it just processes the data.
If I trigger 3 executions immediately:
scheduler.triggerJob(
key,
JobDataMap(
mapOf(
PRODUCT_IDS to listOf(123)
)
)
)
scheduler.triggerJob(
key,
JobDataMap(
mapOf(
PRODUCT_IDS to listOf(456)
)
)
)
scheduler.triggerJob(
key,
JobDataMap(
mapOf(
PRODUCT_IDS to listOf(789)
)
)
)
Then each job will execute with an empty JobDataMap.
So at the end the job's JobDataMap contains only the last element (789).
If I add sleep(100) (enough for the Quartz job to execute) between the triggerJob statements I get the expected result of (123, 456,789).
How can I update the JobDataMap for an existing job without lost update problem?