Crowd Full LDAP Directory Sync running 2 times

[Hans Pesata]

Hi!

Crowd Version: 3.7.0 (Build:#1270 - 2019-10-03)

We are using a Generic Directory Server with LDAP to sync Users from Okta.

We have caching enabled and are using a page size of 200 and it looks like Crowd runs a full sync always 2 times.

It starts with

2019-12-19 15:20:33,569 Caesium-2-3 INFO [atlassian.crowd.directory.DbCachingRemoteDirectory] FULL synchronisation for directory [ 28508161 ] starting

The many entry pairs appear during the sync:

2019-12-19 15:21:35,219 Caesium-2-3 INFO [directory.ldap.monitoring.TimedSupplier] Timed call for search with handler on baseDN: ou=users,dc=xxxxx, dc=okta, dc=com, filter: (organizationalStatus=ACTIVE) took 5863ms
2019-12-19 15:21:35,219 Caesium-2-3 DEBUG [atlassian.crowd.directory.SpringLDAPConnector] Iterating a search result size of: 200

. . .

Then a summary about the number of synced users is displayed

2019-12-19 15:39:45,137 Caesium-2-3 INFO [directory.ldap.cache.RemoteDirectoryCacheRefresher] found [ 36047 ] remote users in [ 1151567 ms ]
. . .

but then crowd seems to restart the sync and the previous entry pairs appear again:

2019-12-19 15:40:08,241 Caesium-2-3 INFO [directory.ldap.monitoring.TimedSupplier] Timed call for search with handler on baseDN: ou=users,dc=xxxxxx, dc=okta, dc=com, filter: (organizationalStatus=ACTIVE) took 8862ms
2019-12-19 15:40:08,241 Caesium-2-3 DEBUG [atlassian.crowd.directory.SpringLDAPConnector] Iterating a search result size of: 200

. . .

until the full sync ends

2019-12-19 16:04:07,625 Caesium-2-3 INFO [atlassian.crowd.directory.DbCachingRemoteDirectory] FULL synchronisation complete for directory [ 28508161 ] in [ 2614056ms ]

The number of entry pairs times 200 is about the double number of user entries.

I wonder what Crowd is doing here by syncing the Users two times ?
Can this be optimized ?

Any hints on what is going on would be greatly appreciated!

Regards,
Hans