In a new preprint, Brešar, Mijatović, and Roberts establish verifiable necessary conditions for central limit theorems (CLTs) of ergodic averages of Markov chains on general state spaces. Their criteria rely on discrete-time L-drift conditions that can be checked using estimates for the one-step transition kernel of the original chain. This yields an integral criterion for when the CLT fails, covering observables such as unbounded functions and indicators of tail events.

The main technical step transfers lower bounds for excursions of the original chain to regenerative excursions of its Nummelin split chain. The same drift conditions also produce lower bounds on return-time tails, modulated excursion functionals, invariant tails, and rates of convergence in f-variation and L^p-Wasserstein distance. These estimates hold without requiring an explicit formula for the invariant distribution.

The results are applied to unadjusted Langevin algorithms, showing preservation of polynomial target tail indices at every fixed step size. For Euler schemes with additive infinite-variance noise, the authors establish restrictions on invariant tails and transience under superlinear drifts. Schemes with linear drift, multiplicative noise, and logarithmic innovation tails exhibit sharp polynomial convergence exponents and exact CLT thresholds, even in the absence of positive Euclidean moments.