Interesting observation on orality vs. writing convention.

I don't know how many different oral recitations Keith heard. I would take his comment to refer primarily to manuscripts.

Yes, but the difference arises purely due to changing transliteration norms. It used to be a quite standard (and probably still accepted) convention to write "~as". In books on Skt grammar by Western authors, especially the older ones, the example for masculine nouns ending in vowel a is usually written as "rAmas", but if you were to get an Indian publication, in an Indian script, only the visarga would be used, thus "rAmaH". Nobody in India would write the letter for s, except in a technical grammatical discussion about the root of the word. I think the older Sanskritists prefer the terminal ~as, simply because that is closer to Greek and Latin word endings.

No. The word is never pronounced as "namas", except when required by sandhi rules. There are standard places within the Rudram where reciters pause to take a breath, and there, if you notice, it is recited as namaH, not namas or namo. I have heard recitations by Tamil, Telugu, Kannada and Marathi speaking people, all of whom have the same uniform standard. The Satara pundit, whose recording you have, would be a good representative of a widespread south Indian tradition. The only thing I can think of is that some reciters have a habit of pronouncing "namonnamaH" with an extra n resulting from unnecessary stress on the na sound. I daresay the careful reciters  regard this as bad form and guard against it.

taittirIya recitation allows one to break sandhi, but this is also quite standardized. For example, in the first verse, ishe tvorje tvA vAyavasthopAyavastha, one can recite ishe tvA Urje tvA vayavastha upAyavastha etc. The pitch values associated with the syllables also get modified in the broken form. The latter is usually resorted to in a teaching context, and the former is used in a ritual recitation context.