Each LN implementation (LND, c-lightning, Eclair and another recent one I can't remember the name of) has its own set of rules and even algorithm to decide which path to follow to go from A to B. And definitely, they have their own rules on how to decide if a node is reliable, or if HTLC's get stuck often, or if response time is too long, or if liquidity is systematically drained, or if one is often offline, etc, so yes, one can work up one's way to be considered reliable.

I think there is a real business in becoming a reliable and profitable node, but most "small" people (they like to call themselves PLEBS) in the field consider it (i) something fun to work on and learn from and (ii) an investment for the future if the LN really takes off. A reasonable goal at present is to try to break-even on hardware and electricity costs.

Do you have more specific questions regarding rebalancing and loop out? Those are vast topics. A good starting point would probably be https://plebnet.wiki/wiki/Main_Page. I'm sure some of the better nodes have their secret formulae, but it's good to start with the basics.

Good luck!