Dating *e > *i in Germanic

Hello list!
I've some observations I'd like to soundcheck with the folks around here more knowledgeable in Germanic history.

1. The Northwest Germanic (NWG) a-umlaut of *u is supposedly blocked by a following nasal. Has any coherent explanation been proposed for this phenomenon?

AFAICT a simple explanation would be to assume not a retention, but a later raising *o > *u, paralleling the raising *e > *i in this same context. However, this seems to suggest that *e > *i was a NWG change, not a Proto-Germanic (PG) one as I have seen usually presented.

Of course, Gothic merged *e and *i unconditionally, so there cannot be any direct evidence for dating this change to PG. Consulting Ringe's _From PIE to PG_ (2006) fails to outline particularly strong arguments, either. A single later change appears to be *VnhC > *V:~hC, which in case of V = *e turns up as /i:/ even in Gothic (and not /e:/). He actually notes that this change "could have spread thru an already differentiated dialect continuum".

This, then, also leaves me suspicious of the number of other changes shifting *e > *i that supposedly operated already in PG times.

2. Unstressed *e > *i. Per Ringe, "it is clear that the ancestor of Gothic first underwent the more limited merger of unstressed *e and *i discussed here, because Gothic exhibits a divergent outcome of unstressed *e before *r, and because this merger was followed by other sound changes that are clearly reflected in Gothic (...)".

The first argument holds no water; this *er > *ar was clearly an _earlier_ change than unstressed *e > *i. Likewise for the change *ew > *aw before a syllable break (which Ringe prefers to analyze as a labialization thru *ow, but where velarization seems to be an option as well; *r and *w trigger similar velarization effects even in later Germanic). The alleged other changes do not seem convincing either, see below.

3. "Early" i-umlaut of *e to *i. It would seem more parsimonious to date this one to the NWG era as well, to group better together all the umlaut changes. The Finnish loanword _teljo_ "boat seat" incidentally seems to show that this change is later than the raising *a: > *o:, which Ringe considers relatively late.

4. *ji > *i after a consonant. Ringe has to postulate here paradigmatic levelling in Gothic, which apparently fails to show evidence for this change.

5. Loss of medial *j. Contractions like *iji > i:, *@ji > ai, *e:ja > e:a are fairly trivial and could easily have occurred in parallel in Gothic and NWG. Gothic also shows *ija > i:, which according to Ringe would've instead occurred later (per some attested relict forms in Old Norse) but clearly seems to be a part of this same sound change bundle.

Are there any other holes in this reanalysis that I am not noticing?

_j.