Glad you found it.

I'd say "won'tfix" for now, the surgery needed to fix this
is probably quite a bit of labor. The modseq based sync is
a huge speed up, so let's try the new code in production first.

Found it. It really has nothing to do with the sync strategy in use. 
The issue is that Kolab/Turba treat distribution lists and contacts as 
essentially two different data types. In Turba_Driver_Kolab::connect, 
each is fetched separately, and following the code down the stack for 
each $type we eventually get to Kolab_Storage_Base::getData(), which 
caches each $type separately. Therefore, even if we already 
synchronized the contact objects, when we query for the 
distribution-lists, this will create (or fetch an already created) a 
separate cache.

This is deep in the existing design of the storage library, not sure 
how to work around that if the two types need to be treated separately.

Ok. Nevermind. I was testing by synchronizing with Outlook. This only 
apparently happens (at least with Turba) when the sync is triggered by 
using the Turba UI to browse an address book.

....or even the kolab topic branch we were using, as it is the same 
code that was merged into master.

Spent some time trying to reproduce this to no avail.

Are you sure you are using the code from Git master?  I only see the 
synchronization taking place once - while watching both the activesync 
log and the general horde log where the KOLAB_STORAGE entries are 
logged. Also, the line numbers and files that the logging is done from 
in your log file don't match up at all and the text of the messages is 
slightly different than from our code.

Hi Michael,

using the new Token sync based strategy, I noticed the initial sync is 
done twice.
Looks like we don't store the initial token / stamp before doing 
another full, initial sync.

---------------------------------
[root@horde-git tmp]# egrep "Full|token" horde.log
2015-07-02T15:50:42+02:00 DEBUG: HORDE [turba] [KOLAB_STORAGE] 
Fetching stamp without token. [pid 25205 on line 248 of 
"/datastore/DEVEL/framework/Kolab_Storage/lib/Horde/Kolab/Storage/Data/Cached.php"]
2015-07-02T15:50:42+02:00 DEBUG: HORDE [turba] [KOLAB_STORAGE] Full 
folder sync details: user: mueller, folder: INBOX/Contacts, uids:   
[pid 25205 on line 248 of 
"/datastore/DEVEL/framework/Kolab_Storage/lib/Horde/Kolab/Storage/Data/Cached.php"]
2015-07-02T15:50:42+02:00 DEBUG: HORDE [turba] [KOLAB_STORAGE] Full 
Horde_History sync for user: mueller, folder: INBOX/Contacts, 
is_reset: 0, prefix: turba:sZ_eVtpyGKcNvfK4lzOlMs6: [pid 25205 on line 
73 of 
"/datastore/DEVEL/framework/Kolab_Storage/lib/Horde/Kolab/Storage/Data/Query/History/Cache.php"]
2015-07-02T15:50:42+02:00 DEBUG: HORDE [turba] [KOLAB_STORAGE] 
Fetching stamp without token. [pid 25205 on line 248 of 
"/datastore/DEVEL/framework/Kolab_Storage/lib/Horde/Kolab/Storage/Data/Cached.php"]
2015-07-02T15:50:42+02:00 DEBUG: HORDE [turba] [KOLAB_STORAGE] Full 
folder sync details: user: mueller, folder: INBOX/Contacts, uids:   
[pid 25205 on line 248 of 
"/datastore/DEVEL/framework/Kolab_Storage/lib/Horde/Kolab/Storage/Data/Cached.php"]
---------------------------------


The full sync request is probably in the code for ages but was never 
noticed before:
The time based sync triggers only every XX seconds and the session 
based sync triggers one per login only.

Personally I don't mind triggering multiple sync requests from the 
specific Kolab drivers right now, we should just make use of the token 
/ stamp if we already have it. Otherwise huge folders will be twice as 
slow to sync as they need to be :)

Attached is a debug log from a newly created user "mueller" + 
completely cleared Horde_Cache.

Cheers,
Thomas