Interesting thread and a very good example why accidental success does not result in true knowledge.
An LED needs a constant current supply, there is no way around that. Thats based on not too complicated and widely published physics. Perhaps that currnt limit can be achieved by some implicit resistance of the components involves, e.g the internal resistance of a battery etc.
However, the question is whether one would want to base a sound design on something that is more a byproduct of a statistically distributed manufacturing process thats only monitored within rlatively wide limits or a design goal kept within close tolerance. A battery for example is a supplier of electrical energy with the usual design goals being a stable voltage over time and mAh etc. The internel resiststance is what you get with it and it may not even be stated on a data sheet.
A resistors design goal, however, is to have a specified resistance and can be bought at different tolerance ratings. Using components for their intended purpose is obviously more a sound approach. That does not mean that you can't get away with not using that approach occasionally. Using that approach, however for a reliable solution is dangerous.