ballpark a 26-0-26 toroid from a 2x35w amp looks right on to be enough to just maintain amplifier about 24vrms into 8ohms x 2 channels at 70-72w total output - it'd just meet the original claim (inspired guesses from experience/how these things are usu put together.). Unless it's exceptionally large, I reckon you are looking at something likely to be right about 100VA. As it is, I expect it'll be a lump what, around 100mmm dia, 30-4mm high? something like that..?
As a result - that's how I'd re-use it / them as a pair as you have: you could wire the two in parallel but better, use one per channel of the amp you want to build, separate rectifiers and reservoir caps per channel. That'll ensure you'll also properly, easily approach double that kind of power output into 4ohms per channel also, because the raw supply will deliver far, far better 'regulation' - lower ripple, and lower sag under transients*.
So -using one trafo per channel and the resultant quieter supply and also shrugging-off low impedance loads,one per channel - is likely much more useful in the real world into real speakers, and domestic listening levels - than stretching for higher peak power into only 8ohms and above. HTH.
* back of envelope - using the two trafos in seriesas one supply to power a pair of channels from higher VDC but single output would potentially have min 8x more ripple at idle; and at least the same 8x or worse 'sag' under sustained transients. No thanks.
This is because- the resulting 'two transformers in series' supply has twice the dc impedance; it is subject to twice the idle current demand / under 4ohm loads, twice the current demand - and there is only the one supply shared between the two channels (= twice the dc impedance, 'seeing' 4 x the current deamnd = 8x teh ripple/transient power 'penalty').
Using lots of reservoir capacitance might be a band-aid, to a small extent, but ..it's a worse solution than just using one trafo per channel, built as a pair of mono amps even if in the same case, and accepting lower 'on-paper peak potential power' output in return for a thing that will actually perform far far better, all of the time.
