If disconnect WS2813 from omega2 pin and connect again - then WS2813 behaves wrong.
Note that WS2813 DIN pins are quite delicate, and very easy to destroy or at least get a latch-up.Never send a data signal to a WS2813 which is not powered with 5V properly. if the cable is longer than a few inches, use a shielded cable for the data. maybe use a serial resistor in the data line to dampen spikes a bit - you have quite heavy undershoots on your scope diagrams you posted earlier in this thread.
First pixel sometimes shifting. For strip consisting of 10 LEDs - in programm we must write 11 LEDs
Sounds like a damaged first LED to me.
I thought it was a problem on my side. But no. It in Omega or yours side.
I'd be surprised.
Because in the meantime, I have used the exact same circuitry for a large installation with 20 Omega2 each driving 518 WS2813, and all of them ran 100% stable severals days uninterrupted. One of these panels still is in my office and I use it to test new stuff every day, and it runs practically 24/7.
Only one caveat: if you have software installed that makes use of the other hardware PWM unit's interrupt, please make sure to use the latest p44-ledchain (OpenWrt 18.06 build here), because there was a bug in handling the PWM interrupts.