Thought I'd chime in here..
The new script is most excellent except for one oversight (can someone bring this to the attention of the script authors?).. There is no way to adjust star detection in the script, but there is a way to do it when you do Star Align on its own. This is a huge problem for those of us with inferior equipment/skills (running Alt/Az and just starting out in AP, so I have lots of very noisy 15s shots of hard to get targets).
Using Star Align directly, I can set it to process all 133 images I have in my stack. Using BatchPreprocessing, nothing flies (I have way too much noise/LP in my frames for the standard detection values).
Secondly, given that there should be a way to parse out any given image (based on various thresholds, and with some kind of "test" button) and decide if it even has usable data (certain percentage of "found stars" are round enough, etc), it shouldn't be too much trouble (for one of you knowledgeable folks) to put in a routine up front of Star Align (or it's own script) that removes files from the file list of light subs to be processed. As noted in other posts, DSS does it quite handily. PI has much better processing once you have frames to process, but the act of excluding images from a stack so people don't have to spend hours going over hundreds of frames to find out if they "look clean enough" should be a no brainer. I mean, isn't the point of computers to speed up what we do with our eyes/hands?.. And clearly it can be done, and with great success, as DSS and others show. --- perhaps another postfix on file names of _g for "good stars"? Then just _g files could be picked for alignment; or another "output directory" option where good files are copied over, or something... anything..
Failing all of that... where is the API info and code/algorithms used today? As a Software Engineer, perhaps I can contribute to PI by making a parser/script to help sort large stacks. I just need access (and some time to learn the code