ANSUZ, the Lost RUNE?

Proto-Norse inscriptions show variously -e and -ai in this position, though mostly -e. Here is what I think happened. Proto-Norse had the so-called ae-to-a change (long vowels), but not Gothic. This change is dated by Voyles to about 200 AD. We think it was somewhat earlier, but there is no hard evidence here. In the absence of inscriptional evidence to date the change, we are thrown back on things like trying to date the Goths' original linguistic separation, etc..
Nonetheless, the change had an important alphabetic implication: the rune used to represent ae was dropped from use, and the usable inventory of runes lowered to 23 (from 24). As a result, the intital vowel of the rune- name also changed from ae to a, rendering it useless it he face of the a-rune (PN *ansuz). Not surprisingly, this is the one Proto-Norse rune for which we have no real evidence as to it's name. I think it was Antonsen who first pointed out the linguistic reason for the loss of this rune. Accordingly, PN had only 5 vowels, each represented by its own rune. Right? Yes and no. Some time after ae-to-a change, and it need not have been long after, unstressed ai was reduced to ae, a sound for which there was no longer a rune.
Now, it is telling that the old rune was not re-adopted. The obvious reason being that it now represented, if anything at all, the same a-sound as *ansuz. We think it likely that the rune only had an honorary alphabetic place at this time, much as certain letters still do in certain modern alphabets.

All of this meant that the new unstressed ae had to be shown by some other means, and the archaic (and etymological) speelling -ai was one of them. The other, and most typical, was -e.
The fact that it ended up as ON -i (through -e) is, of course, irrelevant at this stage. It puts us in a situation where we can still phonetically represent PN accurately using only 23 characters (or their Latin equivalents) , given only that we mark for vowel-length, nasalization of certain long vowels (PGmc. *hanhaz > PN *ha'haz, with compensatory vowel- lengthening, as in Go.), although this need not be shown, and that we use e in unstressed positions to presented ae. This rule is not hard, as ae only occured in unstressed positions in PN (long), while e only occured in stressed positions (variously short or long).

The irony is that while we now understand PN phonology well enough to be able to write it without any problems, we lack the incriptional evidence we need to be able to decline the language without problems, and well as to render its syntax withut problems. We do have problems here, even if they are far fewer that for the PGmc stage. Aside from the syntax, we are really just talking about the correct forms of many of the small words (like the adverb *wel-), but also, and more importantly, a list of case-endings (like the gen.pl of u-stems, the dat.sg. and gen.sg. of fem. i-stems, the dat.sg. of neut a-stems, etc.), and the verdict on the instrumental and locative case uses (where the limited Go. uses should provide some clue). But perhaps PN no longer had any instrumental case except for isolated pronominal- and abverbial-use remnants? This could well be so, but them we would have to explain how the PN masc.neut.instr. sg. ending -u ended up in the dat.sg. of ON strong neut. adjectives, and that is by no means easy to do, especially when we have no inscriptional evidence here, none.

Still, if we hade about 20 more inscriptions, all of them showing a specific form we need (highly unlikely to ever be found), we should be able to hammer out PN with amazing confidence for folk living roughly 1600 years after its last speaker died. Not bad at all, especially when we consider that the language was as close to PGmc. as Go. in regard to changes, but with a syllable-count the same as PGmc. (essentially) .
Lastly, being able to systematically compare a fully accurate (and datable) PN to Go. would go a very long way to solving the problems and riddles of PGmc.