Significantly Slower Sync to Confluence or JIRA in Crowd 2.8 due to /rest/usermanagement/1/group/membership

Crowd 2.8's /rest/usermanagement/1/group/membership is significantly slower than earlier versions of Crowd, 2.7 and 2.6 . Here is an example between Crowd 2.6.5 and Crowd 2.8.0 with 50000 users, and 1000 groups, and 1000000+ group memberships (this is with flat group memberships only. The problem is compounded with Nested Group memberships):

1. Crowd 2.6.5 took 56s to load /rest/usermanagement/1/group/membership

   wget --user crowdperformance --password admin http://localhost:8265/crowd/rest/usermanagement/1/group/membership
   --2015-05-06 18:24:19--  http://localhost:8265/crowd/rest/usermanagement/1/group/membership
   Resolving localhost... 127.0.0.1, ::1
   Connecting to localhost|127.0.0.1|:8265... connected.
   HTTP request sent, awaiting response... 401 Unauthorized
   Connecting to localhost|127.0.0.1|:8265... connected.
   HTTP request sent, awaiting response... 200 OK
   Length: unspecified [application/xml]
   Saving to: `membership.1'
   
       [        <=>                            ] 29,348,141   585K/s   in 56s
   

2. Crowd 2.8.0 took around 8 mins to load /rest/usermanagement/1/group/membership:

   wget --user crowdperformance --password admin http://localhost:8280/crowd/rest/usermanagement/1/group/membership
   --2015-05-06 18:15:32--  http://localhost:8280/crowd/rest/usermanagement/1/group/membership
   Resolving localhost... 127.0.0.1, ::1
   Connecting to localhost|127.0.0.1|:8280... connected.
   HTTP request sent, awaiting response... 401 Unauthorized
   Connecting to localhost|127.0.0.1|:8280... connected.
   HTTP request sent, awaiting response... 200 OK
   Length: unspecified [application/xml]
   Saving to: `membership'
   
       [                       <=>             ] 29,348,141   175K/s   in 8m 2s
   

Both queries return the exact same XML output

Note

This is tested with the same exact user base (via CSV import). It is especially evident with a high user-base .