ablative as ergative marker

[aldai]

Am I right if I assume that unidirectional cross-linguistic paths of
evolution are easier to find for tense-aspect-mood systems than for case
systems? For instance, works like Bybee, Perkins & Pagliuca 1994 propose
so clear paths of evolution for many T-A-M meanings, that it seems they
can be quite warrantly applied to reconstruction. Do we have something of
a kind for case systems? If it is true it is harder to make
generalizations for the evolution paths of case systems, then why? Do we
have to think that cases evolve more multidirectionally? Perhaps following
a general unidirectional principle concrete --> abstract, but not as
straightforwardly as in the case of T-M-A systems? Why so much syncretism?
And so diverse?