Undeciphered languages do not always have a small corpus. For example, the Iberian language of eastern and southern Spain, or the Etruscan language of Italy, have large corpora but even so many aspects of their grammar were unknown (although in the case of Etruscan in the last 40 years much progress has been made, even though the corpus of inscriptions has not been greatly increased).
The reading of both languages is more or less clear (although there is some debate about some signs in Iberian). However, the
understanding of the grammar and syntax is poor, especially in the case of Iberian, where a number of important endings are known, but there is still much debate about what those endings would be marking.
In the case of both Etruscan and Iberian, and also languages possibly related to the former such as Rhaetic or Camunic, and even languages such as Tartessian (which could end up remotely related to Iberian), the problem is not so much the lack of written testimonies as the impossibility of being able to use a closely related language. In such cases, only some form of combinatorial method is possible, but this strategy has not been used massively, because it would require to have the corpus digitized in a convenient way and to use massively combinatorial computation.
As far as I remember, one of the most recent cases of decipherment is that of the Epi-Olmec inscriptions from southeastern Mexico, whose proposed decipherment was based on comparison with the Proto-Mixe-Zoquean language.