Why No shard becomes primary on second node when it is UP again?
Given this post and this one (albeit slightly old),balancing the primary and replica shard across the cluster does not seem to be a priority in an Elastic cluster. As you can see, Elastic sees replica and primary shard and thus the status seems satisfactory for the cluster.
What I would suggest is to have a look at the shard balancing heuristic and play with these values until you obtain a satisfactory result. (as is often the case with ElasticSearch, testing several parameters is what will yield the best configuration given your architectural design choices).
Note that if you start using shard balancing heuristic, you might not get good results if you use at the same time shard filtering or forced awareness