Yea some errors might creep up. Hence periodic diagnostics reports (through email or telegram or slack) are quintessential. Probably have to include more engineering safeguards; triangulate with different data sources. Depending on your portfolio size, you may also want to introduce a self-monitoring architecture across different brokerages - with the reconciliation done at your side.
Anyone Well, I am up to a similar endeavor. While for someone like me who does not have a programming background, it is not that straightforward. And there are a lot to do. It is not only Python, also packages like: numpy, pandas, bokeh, backtesting.py, ib_insync.... Now, I am working on a streaming candle_stick chart, with bokeh. And there does not seem to exist a ready for use package based on bokeh?