What were you trying to do that didn't work?

With a large database the replication online init does not really start because of a scalability issue. The scalability impacts the function 'idl_id_is_in_idlist' that is used to send the entries in proper order.

What is the impact of this issue to you?

The replication agreement only sends the first entry (suffix) then spends too much time building the candidate list.

Please provide the package NVR for which the bug is seen:

It applies on all release but was diagnose on 2.5/2.6

How reproducible is this bug?:

To be detected it require a large database with a large subtree moved under a recent parent.

pstacks are showing RA eating cpu in 'idl_id_is_in_idlist'

Steps to reproduce

will be described later

Expected results

The RA should start sending entries not too long after it sent the initial suffix

Actual results

RA iterates apparently indefinitely without sending any entries