Well, quantum efficiency doesn't change much with bias (having the N/P encroaching on the intrinsic section doesn't really prevent charges from finding their way, there's still a built-in potential). Photocurrent is generally written as independent of bias. This seems to be reasonable, even in the forward direction (PV panels have a squarish V-I curve; basically the only reason current drops is because it's shunted internally by forward I = Is*exp(V/Vth) current). It does do *wonders* for capacitance though!
Heh, nice story. Lesson learned: don't leave nodes floating at undefined voltages! They'll inevitably do something unexpected or undesirable. Hope it wasn't a gotcha, like, from a bad appnote... those are just mean.
Surprised light did much around ICs, unless there was, like, glass body diodes or something. Are PIN diodes usually in glass format? Were those..?
Think I measured this before, a 1N914 is so-and-so, I think around 1nA at room temperature and ambient light levels; rising to 2-4nA under the influence of my (admittedly pretty intense) LED flashlight at point blank range. An effect not attributable to temp rise, because it doesn't vary over time (a valid concern, because I can physically feel the warmth from the LED when turning it on and off near a sensitive region, like my lips).
On the other hand, I've got a circuit with 2N3904/6 collectors driving a capacitor; when both are fully off, leakage is in the ~pA range, for a voltage drift in the 10mV/min range. At least... at room temperature it is.
If you need a really good diode, I've heard a plain old 2N3904 is at least as good as those so-called picoamp diodes (PAD-1 etc.?). Go figure, putting the tiniest bit of effort into making something other than a diode and you get a wonderful diode...
Tim