I doubt that with such a small difference in Vmp that you'll lose much. The MPPT controller should settle on whichever peak in the combined I-V curve yields the most power. As the 18.5V array is powerful and if it has very low volt drops in its wiring then you'll maybe see the lower voltage array stall (as the MPPT controller will settle on a voltage that is above its Vmp and output from that part of the array will collapse). You should have one blocking diode in each series string to stop mis-match losses between parallel strings (for shading and mis-matched Vmp strings like yours). In part light conditions, it probably won't make much difference as the whole array will sit some way below its nominal Vmp anyway.
With my Excel head on... Say the MPPT settles on 34.0V and now your 10x 90Wp panels are making the same 24.32A as before but are operating at 34.0V instead of 37.0V. So your 900Wp array is now 827Wp. But you've gained 13.24A (450Wp) from the new 6x 75Wp array. This is 37.56A solar amps into the MPPT controller and that would translate into 51.0A into a battery sitting at 25.0V (ignoring conversion losses). If you put the 6x 75Wp array on the other non-MPPT controller it will give you 36.0A on the original array into the battery and 13.24A on the new non-MPPT array, making 49.24A into 25V.
The difference will get bigger for a flat battery at 24V and smaller for a nearly full battery at 28V. But most of us want more Amps into a flat battery
So... my money is on putting the new array on the MPPT controller and stomaching the losses on the 900W old array.
My arrays are on two different MPPT controllers because a) one controller couldn't take all the power and b) part of the array is 46.2Vmp and the rest is cobbled together 35Vmp assorted random second hand panels so it made much more sense to use separate controllers optimised for each Vmp