Julia needs code that is type stable, if a function returns multiple types the compiler and runtime can't optimise the calls.
Interesting talk on bootstrapped scaling (a lot of capacity management issues) which touched on audio file distribution and streaming or rather the lack of streaming over HTTP.
Using genetic algorithms in F# to generate monster stats to match the intended higher-level game design intention.
Runs for a fixed number of iterations and asks uses to reassess the goal if a solution can't be found.
Localisation is done on the server and translations are published to the server and the client consumes the language file.
However localised string substitution are done via Aspect programming with interception of setters on objects with the set being overwritten with the static localised string.
OO, you have to love it.
Game design support with Neo4J, game data is checked into git and then a workflow is built on top of it.
Data can be checked and validated and then built and loaded into the datastore.
Design changes are gitflow.