Why Heapster Metric results are minutes lagged behind

Issue

• Querying the heapster metrics endpoint for a pod for memory-usage shows only old data, ~3 min. behind:

$ date
Mon May  2 10:50:52 CEST 2016
[markus@vmapgmstapp4n1 ~]$ /usr/bin/curl -s -X GET -k -H "Authorization: Bearer " -H "Accept: application/json" "https://vmapgmstapp4n1.appad4.tsi-af.de:8443/api/v1/proxy/namespaces/openshift-infra/services/https:heapster:/api/v1/model/namespaces/markus/pods/appagile-php-cassandra-4-2d2vj/metrics/memory-usage"
{
  "metrics": [
   {
    "timestamp": "2016-05-02T10:47:00+02:00",
    "value": 69328896
   },
   {
    "timestamp": "2016-05-02T10:46:00+02:00",
    "value": 71303168
   },
   {
    "timestamp": "2016-05-02T10:45:00+02:00",
    "value": 71303168
   },
   {
    "timestamp": "2016-05-02T10:44:00+02:00",
    "value": 71303168
   }
  ],
  "latestTimestamp": "2016-05-02T10:47:00+02:00"
}

• Is there any way to get more actual data out of heapster ?