How can developers effectively update fresh source code on a blockchain given the challenges posed by the complex and decentralized consensus process?

I am just wondering and struggling to understand how blockchain developers of crypto coins like BTC, ETH, etc. manage to update the code due to the consensus nature. Do they stop all nodes from reaching consensus, so later on push the updated code and then run again and they sync? if yes. How were forks prevented in that case? Or somehow they manage to update the source code without un-plug nodes but instead to have them running up and fully functional?