You are dead right no one understands stats.
The Guardian article doesn't help there as it conflates the shift table with the cluster data.
The shift table data as used in Lucia de Berk's case and correctly criticised by Richard Gill was not used in this way in the Letby case. In the de Berk case it was presented to the jury and they were told there was an odds ratio of 342000000/1 against it being a coincidence that she was the only person on shift. This is misleading given the availability of other explanations for the death and the fact the data provided by the hospitals was not complete.
Primarily because of this issue the prosecution in Letby's case explicitly did not make claims like this. They presented the table as is and invited the jury to draw relevant inferences. At one level it places her at the scene which is critical.
Wider issues with the table, like whether it is complete, are complicated by the fact the CoC classification varied during the period the deaths occurred in so some still births were included in the deaths for that period, hence the varying overall death rate. The defence, of course, could raise these at trial.
But the challenges that arose in the de Berk case, that she worked across hospitals and different grades of unit, don't arise.
The cluster data is much more problematic as this is inherently statistical. The prosecution had to convince the jury the very fact there were more deaths in 2015-16 than in 2013-14 or subsequently was suspicious. Again though, they did not do this by reference to Bayesian probability. They did it by presenting the descriptive stats only and asking the jury to draw their own inferences as to whether it needed explanation and then making the case that the spike in 2015-16 could best be explained by deliberate killing.
This is a deliberate tactic BECAUSE bringing in statisticians has caused misleading evidence to be adduced in the past. the interesting thing about it is that now no statistical analysis is offered and the jury have just been given the data and asked to derive their own inferences (which I completely agree may well be wrong) the statisticians are equally enraged.
Remember when bad statistical evidence has been adduced in the past it has been from people claiming to have relevant expertise. The role of the prosecution is not to be the expert but to hire the expert, and what seems to be annoying the statisticians is 'well yes but we didn't mean no expert we meant a different one'. That's an impossible ask.