I like this idea, but I don't think the source-gathering method is generic enough to work for other cameras. prepCrosstalk uses visit and ccdnum explicitly to retrieve source image data, and it also assumes each detector has two amps named A and B. I thought DECam was a pretty unique situation with all the inter-CCD crosstalk so prepCrosstalk might not be applicable for other cameras anyway.

I'm not recommending that you move your function into ip_isr. I'm just recommending that you add the following to CrosstalkTask:

def prepCrosstalk(self, dataRef):

    return

and change the run signature in CrosstalkTask to:

def run(self, exposure, crosstalkInfo=None):

and add a call to self.crosstalk.prepCrosstalk(dataRef) in IsrTask.readIsrData().

Then you can override prepCrosstalk in obs_decam, but so can others.

Oh I see, this would make a custom readIsrData in obs_decam unnecessary - nice. Thanks for clarifying, will do.

I think I've addressed all your concerns now Simon Krughoff. Please let me know if I missed something. I've rebased on master, squashed some commits, and started a Jenkins run: https://ci.lsst.codes/job/stack-os-matrix/27696/display/redirect. 

I have one small cleanup I'd like to see in ip_isr, but otherwise it all looks good.