You certainly can. But when things go wrong you go from one to three places in which you have to look for the problem - the service, the network, and the server

You certainly can. But when things go wrong you go from one to three places in which you have to look for the problem - the service, the network, and the server. Also, instead of a single package containing a single .exe you're now deploying to all desktops, and to a server. This in the context of developers finish the system, leave, and ops don't have the skills to fix a Ruby/node/whatever tech issue. While none of the options that work are technically wrong, it might be worth considering the differences between the best solution design, and a successful one.