So, if I sum up, when the purge fails:

• The URI used is http://192.168.99.100:9200/com.peopleinsite.account.account/com.peopleinsite.account.Account/_query
• The ES server is indeed 192.168.99.100
• The ES server is indeed reachable on port 9200 using HTTP
• The index com.peopleinsite.account.account is already present on the ES cluter
• The type com.peopleinsite.account.account/com.peopleinsite.account.Account is already present in the index com.peopleinsite.account.account
• The delete-by-query plugin is enabled on every node on the ES cluster
• Other operations on the same index/type seem to work

If all of this is true, then I'm at a loss... It basically means the URI we're using is valid, so one of the assumptions above must be false.

• Do you use some kind of web server between the ES server and your client, which may perform incorrect redirections?
• Do you use any specific configuration on your ES server?
• Are you sure the plugins actually start? I've seen some cases where I had installed plugins and Elasticsearch would simply ignore them. You should have a line like that in your logs if the delete-by-query plugin actually starts:

09:52:35,827 (main)  INFO plugins:180 - [node-1] modules [], plugins [delete-by-query, lang-groovy], sites []