Just had to add one more thing, if possible please add Queue Management preset for maximum upload.
I have 3ghz Core i7 (6/12cores) and 32gb of memory and gigabit ethernet, and with 1000 torrents
constantly superseeded I would like as huge queues as possible. Although Im no expert, but you get
the point, CPU and RAM are almost infinite resource for most, but almost everybody has upload
bandwidth 100% used all the time. So taking everything out of this bandwidth for maximum cumulative
upload is the key, no matter how much CPU or RAM is used for gaining extra cumulative upload speed.

And while I wrote the previous begging, I remembered one point. Have you considered advertising
Tixati for the best P2P solution to 3G/4G and incoming blocked connections? I understand if this
is something you dont want to be profiled, but it could get lots of new users for Tixati and in
P2P in general. Here unlimited data 4G connections are cheap and huge portion of people have only
those connections, and huge portion of those have given up P2P because they dont know or couldnt
find any decently working P2P software.

Im sorry to say, 2.28 didnt solve problems with 4G connections.
Its the best version easily since 2.12, but still it only gets few peer connections
and because of that overall upload performance remains mostly slow. First I thought you had
solved the problem, because it instantly got great upload speed, but that was because
the couple connected peers had enough download capacity. After 20 hour period, every
Tixati 2.28 had 1-20 peer connections, while 2.12 had 100-200. If Tixati 2.28 had
the same amount of peer connections than 2.12, it would be superb.

I have not yet tested this with 100% open ethernet connection, I will report that later.
Im open for testing, anything to help to make Tixati better.

Just checked how Tixati 2.28 worked with my gigabit ethernet connection.
Peer connections and DHT connections were better than with any Tixati since 2.12,
but compared to 2.12 it still is way behind. Peer connections were about 1/3 of
what 2.12 had in those Tixatis that has 1000 seeds and no downloads. Tixati
with 1000 downloads and no seeds had about 75% of connections compared to 2.12.
DHT connections were 1050 to 1100 in every Tixati when in 2.12 everyone has almost
exactly 1200 DHT connections.

As with 4G connection, 2.28 was easily the best one since 2.12, but some change is
still haunting it, while other improvements are narrowing the cap. Anyway, Im open
for any testing you might think would help you to solve this. If you would like to
check how much that auto queue management settings effect to this situation, you could
write custom settings and copypaste them to this thread, where I would copypaste it
to my Tixatis. Anyway, Im excited to help, because I would really like to see how fast
Tixati would be if 2.12's peer connections and 2.28's other improvements would combine.

Yes, this appears to be two different problems, of which one was solved for 2.28 (an issue with super-seeding).  So we're half way there...

Can you turn super-seeding fully off, then try 2.14 alpha-3 and alpha-4, and compare those to 2.12?  With super-seeding off, that will eliminate the problem that was just fixed, and maybe we can then see exactly in which version the unknown problem starts.

I've got a few other problems to solve first, but later this week I'll have another close study of the code differences between versions and see if I can finally get this nailed down.

I tested 2.14a3 and it worked great with superseeding on. It had about 350 peer connections on avarage.
2.14a4 with superseeding on the other hand had only 2-10 peer connections. After turning superseeding off
it had 10-20 peer connections, but that could be just because I left Tixatis running for much longer time
to be 100% sure. But I think it had some effect, because every Tixati had about 3x peer connections.
Still 2.14a4 works like crap. 2.14a3 is as good as 2.12 or could be even slightly better. But as previously
stated, 2.31 is clearly the best one after 2.14a4.

I was away 3 weeks, now I have the time to test whatever, anything to help to fix this.

And this test was done with 50/100mbs 4G connection, with 5 running Tixatis.

I tried to compare uploading of a couple of torrents by uTorrent and Tixati via NAT. In 10+ days uTorrent have uploaded 1.45 Gb and 17Gb, while Tixati have uploaded 3.34 Gb and 1.05 Gb. Default settings. What I can change in settings, to improve upload size?

Im not following your upload size comparison, but Im the original poster of this thread, and I have not found anything
to have much effect. Tixati 2.12 still is the best torrent software for uploading (and I think every Tixati is superb for downloading)
so you should try Tixati 2.12. Compare that to Tixati 2.32 test version, using the same default settings that really are almost optimal
in every situation. If Tixati 2.12 works better for you, please report your Internet connection type, and pretty much what might have
something to do with the situation why newest Tixati is not working optimally for you. And ofcourse the difference with peer connections
and upload speed.

Tixati upload speed is about the same as utorrent when there are less than 20 peers during initial upload of a torrent.

When there are too many peers, Tixati will connect to all the peers in a round robin way.

I try to steer Tixati by changing priority, peer allowance, un-choke clients on seed box peers, but it disconnect them at some point and those settings disappears with time.

utorrent performs better than Tixati when there is a large number of peers because once you reach the Upload slot limit, utorrent stop connecting to new peers.

Correction :
utorrent performs better than Tixati when there is a large number of peers because once you reach the Maximum Number of peers per torrent, it stop connecting to new peers.