Along the southwestern fringes of the Sahara, in the Ennedi and Biltine regions of northeastern Chad and the Darfur region of western Sudan, a few hundred thousand people, the Beri or Zaghawa, speak a language called Beria. Until well into the last century, the Berti people of Darfur and Kordofan still spoke a rather poorly documented related language, Berti; today they are reported to have all shifted to Arabic. Together, they make up the Eastern subgroup of the Saharan family (supposedly part of Nilo-Saharan). I've been looking over some of the literature on these languages lately, so here's a very brief summary on their historical phonology; it's mostly just for my own memory, but if anyone else is interested then great.
Beria is divided into a number of dialects (cf. Wolfe 2001, Anonby & Johnson 2001), of which the best described - thanks to Jakobi and Crass 2004 - is the eastern variety of Kube in Chad. Unfortunately for present purposes, this also seems to be a good candidate for the least phonologically conservative variety. The southeastern Dirong-Guruf varieties preserve /f/, reduced to /h/ in Kube and in the rest of Beria but retained as /f/ in Berti; there is reason to suspect that it was originally *p (for instance, intervocalic variation between /rf/ and /rb/). The western Wegi variety of Darfur preserves intervocalic voiceless stops, which Kube voices, and intervocalic /d/, which Kube merges with *r. There's a lot of cross-dialectal variation within Beria between /m/ and /b/, especially in initial position, which is difficult to account for through regular sound change; word-initially, despite its name, Berti seems to have /m/ in almost all words that have Kube cognates with /b/. Wegi and Dirong appear to preserve a distinction between /l/ and /n/ that has been lost in Kube; but Berti also has /n/ in such cases, so one wonders whether this might be a split rather than a retention, though there's no obvous conditioning factor. It's hard to say much about Berti phonology given the quality of the sources, but it also seems to shift /ɟ/ to [z] in some cases.
Berti is much more closely related to Beria than any other Saharan language, and there are plenty of transparent basic cognates, like "name" (Berti tir, Kube tɪ́r) or "night" (Berti gini, Kube gɪ̀nɪ́ɪ̀). The surprise is that there are also lots of very basic words with no obvious cognates, like the personal pronoun "I" (Berti su, Kube áɪ), or the numeral "one" (Berti sang, Kube nɔ̀kkɔ̀), or the adjective "little" (Berti batti, Kube mɪ̀na). This sort of thing seems to happen a lot in Saharan; maybe more data will make things clearer, or maybe there's a contact context that needs to be better understood. Either way it makes subgroup reconstruction a lot trickier.