In many cases, like in ashleSA, the usage seems to be just due to a mispronounciation of AshleSA (I follow a slightly modified version of the so-called Harvard-Kyoto transliteration in which my capital A is your aa, whereas my lower case a is your single a). The original usage in atharvaveda for the asterism is AshleSA; AshreSA appears in the taittirIYa brAhmaNa (the saMhitA uses AshleSA instead; and atharvaveda does have the masculine AshreSa as the name of an evil spirit). ashleSA does not appear till the mahAbhArata, and is thereafter used only as a name. I do not know of it being used in the a-sleSA sense ever (but then, someone somewhere could certainly have used it).
Note that even mahAbhArata does not follow pANini in places (or, if you prefer, has a lot of ArSa usage), and the Sanskrit language wasn't quite standardized at that time. What seems to have happened is that the -r-/-l- alteration is very old: In fact rlo bhavaH is an old maxim. It is also seen in other parts of the Indoeuropean family of languages. But, the beginning syllable was definitely an open long A- at this earliest stage (stressed when used with a verb, unstressed when the suffix usurps the stress). Somewhere before/around the beginning of the common era, the name asleSA seems to have appeared (and around this time, stress disappeared except in ritual contexts).
The original A- is clearly the prefix A- (which appears in Agama, AmantraNa etc.) which in Vedic was a separate indeclinable particle, but like in German untrennbar verbs became an integral part of the verb AshliS in classical Sanskrit (except, of course, the perfective/past marker a- still slips between it and the verb.)