It’s not that simple. Taking a quick look at their development teams it appears that they are mainly coding in Java with a sizeable arm in Bulgaria. One thing they could do is try and refactor the could into C++. But finding the amount of specialist low-latency developers with Java/C++ in Bulgaria would be a challenge as, whilst Bulgaria has a reputation for high coding standards, it is still a smaller EU country and a lot of technical talent is taken out of the country or employed remotely at above market rates. Refactoring the code would take a long time, a project over many years.
They may be confident they can still get better latency times via Java, we don’t know how far through the development projects they are. But again, improving the latency isn’t a case of just pressing a few buttons. It could be a case of many months before deployments add up to significant change.
It is evident that they are using microservices and multithreading/concurrency (which is expected) so the basic architecture looks good. But it may be that there is an existing monolith there that needs to be scaled out before it can be decoupled and split up, so they can reduce load on the system. This would take time. Some monoliths are huge and complex and take years to fully break up.
Hard to tell what is happening on the GUI side from looking at the development teams. But their work will be tied to what the core system can facilitate.
The simple truth is that a certain level of patience will be needed from us as they try and meet growing demand for their product. So we need to invest with that in mind.