According to that datasheet the only thing that can put DC on the output is the input. If the input is capacitively coupled it would seem to have to b e a defective IC but you have changed it which pretty much rules that out. So now what ?
I say when shit like this happens look at the PC board. If there are no set ting resistors for the DC level on the input, any leakage path will do it.
I called Jethro bodine to cypher this out and we got 26 dB gain. well 20 dB is about 100, plus 6 makes it about 400. (approximately, fuckya about prec ision). The input impedance is 60 K, and since it a semiconductor and we ai n't talking Mhz that means the input resistance is 60 K.
Let's get conservative on the gain. It is say, 200. That means for a six vo lt output it would take what ? You have 14 volts on the board, a leakage re sistance of six megohms could result in 1/whatever at the input. Multiply i t back up by the gain and you got volts, not millivolts.
Such leakage paths have many causes. Sometimes customers spill something in it like soda, and after years it starts becoming conductive. Another big o ne is electrolyte leaking from the caps. This is another insidious one beca use sometimes you cannot see it and it (or the soda, or bigscreen CRT coola nt) gets more conductive as heat and voltage is applied, and sometimes it r everses when turned and cooled off.
Tell ya what, take a 10 K resaistor between pins 1 & 2 of that chip and see what happens.
Either that, or take the thing apart and clean the PC board with acetome an d alcohol. (acetone first, then HOT water, and then alcohol)
If you replaced the coupling caps, it has to be some kind of leakage. there is no bootstrap or any of that shit here. This is simply two OPAMPs with c ommon feedback, there ain't no more.