The only difference that I could find was that the
alkalines had a cell
voltage of around 1.65-1.7 volts, feeding the clock with something
nearer to 7 volts. I have no idea why this upset it, but the battery
voltage circuit reported this as flat batteries.
I actually have a PC2086 serice manual somewhere.(Amstrad service manuals
of that date were just parts lists and schematics, of course.) From what I
remember there's noting odd about the circuitry, though.
Did you try a couple of silicon diodes in series, in series with the +ve
battery lead? That might have cured it. I have an objection to
zinc-carbon cells due to their antisocial habit of leaking, alkaline
cells are much better house-trained, so I prefer to use those where possible.
-tony