Issue with lookups cached global caching behaviour

Hi,
we have a couple of lookup definitions and specify a poll period of 3 hours but observe that our brokers are actually reloading the lookups every hour and the historicals are even reloading them every 30 minutes.
To use one of the lookup definitions as an example, the following is what the coordinator endpiont “v1/lookups/__default/adspace_by_adspaceid” returns:

{
  "type": "cachedNamespace",
  "extractionNamespace": {
    "type": "uri",
    "uriPrefix": "s3://XXXXXX/adspace/",
    "fileRegex": ".*V2.*",
    "namespaceParseSpec": {
      "format": "tsv",
      "delimiter": "|",
      "columns": [
        "AdspaceId",
        "AdspaceName",
..........................
      ],
      "keyColumn": "AdspaceId",
      "valueColumn": "AdspaceName"
    },
    **"pollPeriod": "PT3H"**
  },
  "firstCacheTimeout": 0
}
I also checked the lookup snaphot file on one of hour historicals and see the same defintion there:
  {
    "factory": {
      "type": "cachedNamespace",
      "extractionNamespace": {
        "type": "uri",
        "uri": null,
        "uriPrefix": "s3://XXXXXXX/adspace/",
        "fileRegex": ".*V2.*",
        "namespaceParseSpec": {
          "format": "tsv",
          "columns": [
            "AdspaceId",
            "AdspaceName",
..........
          ],
          "delimiter": "|",
          "listDelimiter": null,
          "keyColumn": "AdspaceId",
          "valueColumn": "AdspaceName"
        },
 **        "pollPeriod": "PT3H"**
      },
      "firstCacheTimeout": 0,
      "injective": false
    },
    **"name": "adspace_by_adspaceid"**
  },

We have 28 namespaced lookups and each of them has the same PT3H poll period.

However, in the historical.log file on the same historical node, I see around 28 loading and deletion activities every 30 minutes:
2017-05-26T04:03:11,518 INFO [NamespaceExtractionCacheManager-0] io.druid.server.lookup.namespace.URIExtractionNamespaceCacheFactory - Finished loading 279487 lines for namespace [d90524b3-e949-4b03-baad-12b918ff5bbd]
......
2017-05-26T04:33:05,175 INFO [qtp1127856876-56] io.druid.server.lookup.namespace.cache.NamespaceExtractionCacheManager - Deleting namespace [d90524b3-e949-4b03-baad-12b918ff5bbd]

Every time, those unique identifiers are different. I also had it once that after a QTL lookup update by posting a new definition to the coordinator, one of our nodes started reloading the lookups every minute.
After clearing all lookups in the metastore and reposting them via the coordinator, all went back to normal.