If we don't use STC at all, then wouldn't it make more sense to set the microtime onset to the slice that represents t0, in your example (ascending interleaved), slice #2, which is the first slice being acquired? (so that it aligns with the event times).
Well, if you don't go with STC at all, then most of the slices are slightly shifted in time. Assuming all the data of a volume to be acquired in the middle of the TR should lead to the least misalignment (- TR/2 for the first slice, ... + TR/2 for the last one, if you sum up the absolute values this is smaller than 0 for the first slice, ... + TR for the last one).
Sorry for the email onslaught. Has this changed in SPM12? In my case, I have 36 slices acquired in ascending order. I do NOT perform slice timing correction. Like the original poster, I thought that it was best for me to set microtime resolution = 36 and microtime onset = 1. After seeing this post, I re-ran with microtime resolution = 36, microtime onset = 18. However, the former analyses actually gives me substantially more signal.