Grepping for [^<-](\ba\s+[aoei]\w+) revealed many more occurrences of "a" article typos.

Grepping what revealed more occurances? Your grep pattern returns nothing for me in the source tree.

$ grep -e '[^<-](\ba\s+[aoei]\w+)' *.c
<no output>

The OP's expression uses PCRE features. The GNU and BSD grep implementations accept \b and \w under -E, but not under -e. Or, you can use full-fat PCRE in GNU grep with -P.

I've fixed all the ones in www and src.

The regex still matches some stuff, but I think they're all kosher. The biggest offender is "…a one-time…", but I can only speculate why English works this way. (It isn't a vowel in this case, it's a "W" sound.) Sorry, world; the language was like this when we got here.

Now, happily off-SCM-topic: this comes from a non-native user, but with a keen interest in languages and speaking a handful of them too.

The nice theory handbooks tell that a vs. an is not about the next word beginning with a vowel letter, but with a vowel sound. As is, this is quite easy for the usual suspects like a and i, but less easy for the masters of camouflage like u or o:

a university / an umbrella

a one-time chance / an oven

not to mention the favourite chameleon of the Indo-European languages, h:

a history of Fossil / a[n] historic SCM / a horse

So, Warren is right to acquit "one-time" and spare it the parole officer "n".

While not all of language was indeed there when we got here (I happened to witness an average of one official spelling reform per spoken language, and several dictionary-sanctioned changes in the "preferred form" of quite a few words), the a/an rule in English looks pretty stable.

I always found this an good way to spot non-English posters, as they often get it wrong -- in both the "too little" or the "too much" directions ;-)

It's been an honor to read your reply.