nothing. Most important thing is data quality, data diversity, and an easy to use sdk/api. All the other flashy UI things are never used by anyone other than someone who can't really use the data
for python, no threads. Subscribe to all tickers you want and handle the callback if needed in a separate worker thread (which probably is not needed). ib limits the rate of market data anyway so a separate thread/process is useless.
50 price points/second: python can handle this easily. I bet...
if you already have the current moving average you do not need to keep all the old values. Imagine if your window was in the billions, your buffer would be huge. you can subtract the mean and add the new value
Microsoft office is used alot. The OS is used by almost everyone in every company. Azure is ok. Visual studio and dot net are good.
In the 90s some firms tried solaris and neXT front to back but either failed or went to windows for front
infrastructure, credit lines, access to multitude of products, reduced fees. Some firms provide so much that you never hear of employees leaving and starting their own firm