Depends. I don't work much on data analytics these days so only a few of the systems I've put in place (using SQLite + bash scripts on a workstation) are still in use. Most have been shut down for organizational/business reasons. They did their job and typically the entire project was put out to pasture, not upgraded to other technology. Further reaffirming my suspicion that investing in such complexity up front was not warranted.

The biggest risk is that your software is successful and you're stuck with a ball of hacks that no one except you knows how to run or debug! Good documentation and automation can keep it alive, but not indefinitely. If your project is successful, you can expect it to get more complex by default.