groovy - Why are my Elasticsearch shards sporadically failing when running queries against my alias filter? -

i running elasticsearch v.1.6 setup 2 nodes, , index 5 shards. index, have added alias contains reference index, groovy script filter helps avoiding documents invalid start , end dates. alias looks this:

"myindex" : {    "aliases": {         "myindex_alias": {              "filter" : {                 "script":{                      "script":                         "\r\n entrydate = doc['availability.start'].date.getmillis();                          \r\n exitdate =  doc['availability.end'].date.getmillis();                          \r\n rightnow = datemidnight.now().getmillis();                          \r\n\r\n if (entrydate < rightnow && exitdate > rightnow)\r\n {                                        \r\n return true;                              \r\n }                              \r\n return false;",                         "lang":"groovy"}}}}} 

when performing simple query, such term * (/myindex_alias/my_type/_search?request.q=*), number of shards fail query returning following message:

"failed": 1, "failures": [     {         "index": "myindex",         "shard": 2,         "status": 500,         "reason": "remotetransportexception[[myescluster][inet[/10.131.43.61:9301]][indices:data/read/search[phase/query]]]; nested: queryphaseexecutionexception[[myindex][2]: query[constantscore(scriptfilter(         entrydate = doc['availability.start'].date.getmillis();         exitdate = doc['availability.end'].date.getmillis();         rightnow = datemidnight.now().getmillis();          if (entrydate < rightnow && exitdate > rightnow)         {              return true;         }         return false;))],from[0],size[10]: query failed [failed execute main query]]; nested: groovyscriptexecutionexception[arrayindexoutofboundsexception[null]]; "     } ] 

i have verified following things:

1. all documents in index have valid start , end values, , none of them have null value.
2. the filter does work on shards perform query successfully
3. the cluster has green health status
4. the error occurs randomly on shards both in master , worker nodes
5. the errors not occur if queries directed index instead of alias
6. adding more nodes cluster seems mitigate problem. running 2 nodes leaves 2-3 shards failing, while 3 nodes limits failure 1-2 shards. adding final 4th node leaves query failing on single shard
7. even when query succeeds, fewer results expected, again sporadically

any on matter appreciated!