Without digging into your configuration, I can say that the most likely cause is that your mangle tables aren't creating connection tracking entries for route marks on new connections originating on the various WAN interfaces. That's the most common mistake I've seen in posts with your problem.

ok - it looks like your NAT rules are to blame. You're using netmap which is a stateless nat action - that means you must use TWO rules to accomplish each mapping.

I see why you thought to use this, as it's apparent that you have a 1:1 relationship between a specific public IP address and private IP address, yet you're doing all of these mappings at the protocol/port level. I recommend against this methodology. If you want to dedicate an IP to a 1:1 mapping, then you should just map it at layer 3 (the entire IP address, not just certain ports) and then limit access to the host by using filter rules. This ends up being cleaner, easier to understand, uses each section of the firewall for its intended purpose, and takes less CPU.

So let's take the host that is 178.76.252.198 -> 192.168.0.90
You have the following rules:
add action=netmap chain=dstnat dst-address=178.76.252.198 dst-port=443 in-interface=1-TTK protocol=tcp to-addresses=192.168.0.90 to-ports=443
add action=netmap chain=dstnat dst-address=178.76.252.198 dst-port=443 in-interface=1-TTK protocol=udp to-addresses=192.168.0.90 to-ports=443
add action=netmap chain=dstnat dst-address=178.76.252.198 dst-port=5061 in-interface=1-TTK protocol=tcp to-addresses=192.168.0.90 to-ports=5061
add action=dst-nat chain=dstnat dst-address=178.76.252.198 to-addresses=192.168.0.90

I'd say the best thing for this host would be to just remove the netmap rules. There is no "return-path" mapping to undo the stateless nat, so anything getting translated by these rules will not be properly "un-natted" on the return path. The final rule here actually does exactly what I'm recommending (map the whole IP and use the filter table to limit what actually gets forwarded).

Finally, your route rules seem to be a bit confusing to me:
/ip route rule
add src-address=178.76.252.192/29 table=to_1-TTK
add src-address=109.195.238.0/24 table=to_2-DOM-RU
add dst-address=192.168.0.0/24 table=main
add dst-address=192.168.2.0/24 table=main
add dst-address=0.0.0.0/0 table=main
add routing-mark=to_1-TTK table=to_1-TTK
add routing-mark=to_2-DOM-RU table=to_2-DOM-RU

All of the underlined rules don't seem to make sense to me, especially the 0.0.0.0/0 rule (unless you're temporarily using this to disable load balancing for troubleshooting purposes)