This is the explanation of the failure case:
"If the most recent delivery result was a failure, then no more attempts will be made for a period of time.  Initially, and on the first failure following a success, this delay will be DeliveryQueueHelper.getInitFailureTimer() (milliseconds).
If, after this delay, additional failures occur, each failure will multiply the delay by DeliveryQueueHelper.getFailureBackoff() up to a maximum delay specified by DeliveryQueueHelper.getMaxFailureTimer().
Note that this behavior applies to the delivery queue as a whole and not to individual files in the queue."

In case of subscriber not available multible times the retry-timeout seems to be to high. We had sometimes to wait for more than 15min that the delivery to a subscriber worked again.
This retry timeout should be put to lower value.

To update the values we can use the "internal" api

curl --location --request PUT 'http://{dr-prov-url}/internal/api/DELIVERY_FILE_PROCESS_INTERVAL?val=60'

curl --location --request PUT 'http://{dr-prov-url}/internal/api/DELIVERY_MAX_RETRY_INTERVAL?val=600'

dr-node will pick up its config automatically but to trigger manually

curl --location --request GET 'http://{dr-node-url}/internal/fetchProv