1. In aorist vs. stem forms of verbs, when the initial consonant of the aorist form is glottalized, it is not glottalized in the non-aorist form. We've already seen this with t'omom-, "to kill (aorist)", vs. doom-.
2. In aorist vs. stem forms of verbs, when the medial consonant of the aorist form is glottalized, it is not glottalized in the non-aorist form. I previously thought that might just have been a case of an impermissible consonant cluster. We previously saw this with mats'ag- vs masg-, "to put".
3. In aorist vs. stem forms of verbs where the non-aorist form has a glottalized final consonant, which is not glottalized in the aorist form. Sapir gives to examples which are not apparently triliteral roots:
"drown" stem: nuut'-, aorist nuud-
"spread" stem: wiik'-, aorist wiig-
As with vowel ablaut, there are also cases where consonant ablaut is not connected with a regular rule, but I don't understand Sapir's examples. One of them is:
somod-, "boil"; ts'umuumta- "boil"
In this case, the second form looks like what Sapir elsewhere calls the aorist frequentative with some kind of -ta suffix. My initial guess from somod- would be that it is an aorist form, since otherwise I would have expected somd-. I suppose that is where the mystery lies: why does the aorist frequentative differ from the aorist by having both raised vowels and a glottalized initial.
One possible explanation could be that the aorist form somod- is a back-formation from non-aorist *somd-...but we would need some more evidence before we could know for sure.