This bug fix (https://fedorahosted.org/389/ticket/47942) was already reviewed by Mark.

Taking into account a first remark, I change the indentation of the previous fix (IDE are sometime teasing friends :-) ).

The second remark was that the same bug applies both to incremental and total update. The fix above addressed total update.
The fix below addresses the incremental update.

In that last patch, window/pause are hardcoded (100/2s). I was unsure if the adding new configuration attribute was valuable.
The consequences of this bug for incremental update is less critical than for total update. With incremental, the RA will hang for idletimout but then a next replication session will continue sending updates. While for total update, this bug is fatal.
It is also less likely to occurs (amount of transfert data (update/entries), job to do on consumer side to apply the update, #update to send (#update < #entries).


thanks
thierry