Sorry Sarasir but I do not understand anything what you have done with analog front end and why.
I have inspected many versions of Owon analog front and I have not finded any reason for modify shieldings etc (exept first version where BNC's was not at all instelled to PCB). Front end is one of best front end in this class if look example its noise level and frequency band. Also Owon input capacitance is originally typically quite low. It was this layout version we measure it by VNA and it was "around" 8pF (under Owon specs). There is not any reason for try get it more low. Tektronix 2465 have 15pF and they really know what they do. Same in old HP, in this case 8pF in 500MHz scope and agen in one Tektronix 500MHz it is 10pF. (nominal +/- something)
In original condition input circuits do not picup any markable amount of internal or external noise around it. Close differential input (50ohm terminator is enough and zero z do not make big difference) and you know it after you measure noise figure and repeat it with open differential input (example using 50ohm terminator without center tab connected) and measure noise figure (not only level). In this case differential input other pole is exactly this part of BNC outer terminal what is just in position where is its inner terminal (connection point). This is reference point and nothing else if think oscilloscope input.
Compare this noise figure with Agilent, Tektronix, or even Rigol better models than 1052.
Now you tell that your input stage oscillate? (really? )
You have changed parasitic components?
Just return all exatly as it was after factory (I mean only front end - but overall nice experimental work
specially for reduce SMPS generated RF noises) and you have reasonable good front end well comparable with Agilent and Tektronix much more expensive models. In worst case, modifying shielding there may lead even worse result even if it looks like "better" physically with eye. (it is due to fact that where these noise currents flow) (this is not for Owon but sometimes example changing RF field "GND" connection points may lead worse situation because it change noise current road. It is good to imagine that there is not common homogenous GND! Every single point in GND is different where ever it is. DC and low audio is different and 1MHz is quite easy but 1GHz is more difficult and 1THz is just magic.
In Owon - just analog front and also ADC is very good. Other good part is very good quality TFT.
added sidenote:
it looks that your front end version is version what can not fully adjust without separating front panel board and main board. Some version can adjust more easy.