The subject is something of a musicological can o' worms.

To wit:
1. No one REALLY knows how Latin was pronounced when it was a "living language."
1a. Some people think they know. They are wrong. We can't know. We can only guess.
1b. The idea that a language was pronounced one particular way all over an Empire spanning portions of three continents is ludicrous anyway. People in different parts of Germany TODAY pronounce their own shared, native, living language differently.

2. The standard "Church Latin" pronunciation is, essentially, the way Latin was being pronounced by Italians at some particular time in the past.

3. It is commonly assumed (probably rightly) that German-speakers of the time of Bach were pronouncing Latin some particular way. See (1a) on the danger here, and also (1b) on why it would be different than described in (2).

4. It is considered good form by some (but not all) for music written in Latin by Bach (and his German contemporaries) to be performed using the assumptions described in (3).

5. Among those who subscribe to the concept described in (4), there is disagreement about the particulars of (3), a problem explained in (1a) and further developed in (1b).

6. The only people who need to know about, or have an opinion on (3), (4), and (5), already do.

7. If you are not one of the people described in (6), you should probably stick to the pronunciation described in (2) whenever required to be the one pronouncing.

The most important thing is that everyone in the choir do the same thing. Conductor gets to make the call.

Isn't it common practice in the U.S. to follow Italian-style Church Latin pronunciation in Mass, regardless of the composer's nationality? Concerts are another matter.

In Austria there is a monastery named Heiligenkreuz, and a month ago the Sunday Chant Mass was broadcast on German TV. They too pronounced the chants in the German way.

I'll try to explain the way Latin is usually pronounced in Germany using IPA:

a = [a]
e = [e] or [ɛ] for long or short vowels
i = [i] or [ɪ]
o = [o] or [ɔ]
u = [u] or [ʊ]
y = [y] or [i], depending on local custom; Austria seems to prefer [i], Germany [y]
æ = [æ]
œ = [œ] or [ø]

h = [h]
j = [j]
v = [v]
q = [kv]
r = [r], [ʀ] or [ʁ] depending on local custom; never as [ɹ]
s = [z] as initial or medial consonant, [s] as final consonant or before other consonants; there is some liberty in using [s] also initial.
z = [ts]
c = [ts] before a, e, i, æ, œ, y; [k] in other instances
cc = [ts] (sometimes [kts]) before a, e, i, æ, œ, y; [k] otherwise
ch = [χ] or final [ç]; for foreign words from Greek
th = [t], for Greek words
ph = [f], for Greek words

Therefore we have
sanctus = ['zaŋktʊs], crucifíxus = [kru:tsi'fɪksʊs], magníficat = [mag'ni:fikat]

I just discovered there is a good page on Wikipedia on this topic, Latin regional pronunciation.

Here is a video with clips of Vespers of Pentecost at Heiligenkreuz (unfortunately not complete, in the beginning the first two syllabes of [Domi]ne, ad adjuvandum me festina are missing and some similar occurrences).

The b-minor Mass was written for the catholic court of Dresden, and short Masses (i.e. only Kyrie and Gloria; Lutherans have another understanding of Missa brevis than Catholics) were used frequently used in the Lutheran church in those days. Leipzig was and is a university city, and the use of Latin well into the 18th century is well known. He also composed Magnificat settings in Latin and I suppose he used them for Vespers. Buxtehude was also a Lutheran and composed some Missae breves, the well-known Membra Jesu nostri and some Magnificats in Latin.

I thought I read once that Roman Latin was particular law for the Church in the US. That was doubtless pre- V2, but I can't imagine it would have been changed (why legislate on what's not in use?). Can anyone confirm that? It was certainly encouraged by Pius X.

If this is in fact the case, then German Latin is only of use in concert performances. German choirs have tended to use German Latin regardless of provenance of the music.

I'm open to period pronunciation in early music performance, but I have to admit that Franco-Flemish Latin ca. 1500 drives me around the bend. "Credo in EWWnum DeEWm"

I had a voice teacher who had no qualms about changing the vowel (or the entire word for that matter) if it "sang better" for the voice. So much for the text being subservient to the music. But in her defense she was an opera singer and not a liturgical choir director.

For a long time, Latin was pronounced as if it were the language of the country. This is set out in great detail in the book "Singing Early Music," ed. Timothy McGee (Indiana University Press). In fact, this book is only about the historical pronunciation of vernacular languages and Latin for use by musicians performing early music.

In the seventeenth century Christoph Bernhard said that it was acceptable to pronounce it as the language of the country, but it was better to use Roman Latin. Prosper Guerranger, the first abbot of Solesmes, campaigned for French churches to use both the Roman Missal (the French often had their own diocesan missals) and Roman pronunciation, i.e., church Latin.

The historical pronunciation is very useful in the performance of polyphonic music, since the sonority of the language is a crucial element in the sonority of the composition. But for liturgical usage, I always use church Latin. In the performance of polyphonic music, though, the vowels can be colored somewhat to achieve something of the sonority the language gives the music, without departing from the parameters of church Latin.

Singing in Latin by Harold Copeman

That's the book which was escaping me earlier in the thread.