CPU: Exynos 5422 Octa big.LITTLE ARM Cortex-A15 @ 2.0 GHz quad-core and Cortex-A7 quad-core CPUs
Memory: 2 GB LPDDR3 RAM at 933 MHz (14.9 GB/s memory bandwidth) PoP stacked
This could be because
You have several users opening analytical type dashboards at the same times (where they're just requesting graph after graph while using it).
You have a very complex control algorithm / predictive analytics
You need to run large reports
Past that, an ES, having less computing power than the prospective powerful computer, would be less forgiving with inefficiencies created in configuration, like
Not using Websockets or the AngularJS dashboarding tools and directives for dashboards.
Inefficient meta or scripting data source scripts.
To the root of the question, we usually would recommend adding another Mango (perhaps in the cloud) and then move the analytics and display responsibility onto that instance (so, publish all the data from the ES). Then the ES is certainly able to handle thousands of data points. It divides the user access for better distribution of the responsibilities of the system (leaving control unburdened with servicing user requests typically).
This forum is not the right place for your general purpose computing hardware questions. Those places exist, your question belongs there. One of your links appeared to be SEO spam, which will get you banned if you do it again.
Yes, quite possible. You could get started with a Serial Data Source. You'll need to identify the serial port, find the port settings in the device documentation (baud rate, parity, etc), and set up the regex to capture in incoming data.