> 1a. Are there satem languages that didn't lose the /n/ in their word for 100?
the /n/ in the word for "hundred" in Proto-Indo-European (PIE) was a vocalic nasal /m̥/, and is reconstructed as *ḱm̥tóm.
PIE Sanskrit Old Persian Latin English
--- -------- ----------- ----- -------
\*ḱm̥tóm śatám satem centum "hundred"
In Indo-European vowel gradation, this is known as the zero-grade form [1], and its phonetic realization evolved differently in daughter languages. /m̥/ became the -en' in Latin "centum" (the /n/ is due to assimilation with the following /t/, and "a" (schwa) in Sanskrit "śatám" (Old Persian "satem")
Furthermore, this transformation occurred in all instances of the zero-grade vocalic nasal, regardless of whether it is adjacent to a PIE *ḱ.
For another example, PIE *gʷmsḱeti ("goes/goeth") becomes Sanskrit "gacchati". Notice that in this case, the /m̥/ transforms to "a", just like in *ḱm̥tóm -> śatám, but here after *gʷ.
This demonstrates that the transformation is independent of whether the preceding velar sound is palatalized or labialized.
The original nasal /m/ is however retained in other forms of this Sanskrit verb, like its root form "gam-" (PIE: *gʷem-) which is in the lengthened grade [2].
> 1b. Are there centum languages that merged k into ḱ rather than the other way around?
None that I'm aware of, but happy to learn of an example.
> Why does the geography of Tocharian matter to the hypothesis that centum / satem reflects an early split between "proto-centum" and "proto-satem"? People move;
You're right that geography isn't deterministic, especially as it pertains to language, and Indo European has been a very mobile language family.
That said, most of the centum languages to this day are western languages and most of the satem are eastern. The big exceptions are Slavic (satem), Mittani (satem), and Tocharian (centum).
1. https://en.wikipedia.org/wiki/Indo-European_ablaut#Zero_grad...
2. https://en.wiktionary.org/wiki/%E0%A4%97%E0%A4%9A%E0%A5%8D%E...