Here's diff for the fixed and rebased patch. There're some unexpected things concerned with the check of emptiness of destination table: internals can only correctly deal with ABORT conflict action and also my 'small optimization of optimization' was incorrect.

пн, 16 июл. 2018 г. в 16:20, Nikita Tatunov <hollow653@gmail.com>:

пн, 16 июл. 2018 г. в 16:06, n.pettik <korablev@tarantool.org>:
Could you please rebase patch on fresh 2.0?
After that, I will give final assessment.

Yes, no problem.

Also, did you manage to check that xfer optimisation really works at runtime?

Yes, I have it in newer tests and involved an int variable on this purpose. Only need to make small fixes in it.