Robustness on lima node decoding

This MR is a hotfix applied on ID10.

This problem was impacting both ID15 (on which redis data base was cleared) and ID10 (where this patch were applied).

The exception was failing the BLISS data viewer and Flint at start up, after few seconds. As result there was no way to display data.

I guess it means something wrong was stored in Redis? Or it is not the way to decode stuff. Is somebody could review seriously this code? It would be good to know how this can be reproduced, and how we should fix the problem.

@matias.guijarro @denolf @sebastien.petitdemange if you have an idea.

Related to #2128 (closed) (mitigate the problem for Lima nodes)