The determining factor for your design is LED forward voltage (Vf) spread.
If all have identical Vf and you drive your series-parallel string with constant current, you will have equal current in each string.
The world is however not perfect, so there will be natural variations in Vf between LEDs, especially identical LEDs for different batches.
Let's take Seoul S1W0-3030XX8003 for example:
Within the same bin, you need to assume 100 mV difference between individual LEDs. If you are mass producing them and don't buy them binned at all, you need to assume the full 200 mV spread. Within the same reel of LEDs, you should have the same bin, save for some really shady manufacturers. But if your EMS does not guarantee that all LEDs on the same panel is from the same reel, you can't assume it.
So how much does 100 mV matter?
There are redeeming factors. The LEDs are not ideal in the sense that they only have forward voltage with an infinitely sharp I-V knee but rather equivalent resistance, causing a slope. For an infinite parallel array with all but one being 2.8 V and one is 2.7 V, the 2.7 V one will see 2.3 times higher current than the others. You can design for this, but it means significantly underdriving your LEDs, causing bad economy since you need to buy more LEDs.
Your array isn't infinitely parallel, so it won't be that bad. Your 2.7 V one will take more current than the 2.8 V ones, but less than 2.3 times. Please simulate your exact series-parallel situation with one sting all having the lowest Vf rated and all the others the highest. Your model needs to include the equivalent resistance of the LED to give the correct I-V characteristic.
Another factor is that the LEDs with the lowest Vf that will see higher current will also heat up more, and the white LEDs have negative temperature coefficient, worsening the situation. Red LEDs on the other hand have positive tempco.
Your best solution for efficiency is to run them all in series. It is however the least fault tolerant solution and one LED going open will turn off everything. A Zener diode with higher voltage than your max Vf across each LED is one way to mitigate it, but there is a static power loss penalty from it since they will leak a little bit of current.
If you are stuck with a series-parallel string like yours, you need to either straight up accept the result from max spread within your Vf bin(s). Tighter Vf spread = less overcurrent needs to be derated for. If unacceptably high, the simpleast but least efficient remedy is a resistor for each string. If you have small production quantity, you can hand-match just the low Vf strings when they appear and add just enough resistance to balance out and with zero ohm for the high(est) Vf one(s). With high volume production, this won't be an option and you have to design for worst case scenario you can accept and suffer from that even your string with highest Vf has the same resistor in series as your lowest Vf one. An intermediate option is to have an active device such as a linear constant current sink for all strings but one. That will come with its own set of tolerances but will improve your situation.