It looks as if you are already using some sort of method to figure out how language is used. In fact the use of good dictionaries is probably the best thing to start with, but let me mention some other ways language can be examined. What I am referring to, specifically, is the linguistic term "collocations" and the many databases out there that help one identify and study them. Granted, each language has its own patterns and those that have been identified can help a language learner immensely. For example, knowing that German one-syllable words ending in the letter "z" tend to be masculine is rather helpful. Short of that, what we have are random collocations that, short of massive exposure to the language, will simply have to be memorized. On that note, I'd like to point out some very helpful collocation databases you may or may not already be aware of. I will use the prepositions you've asked about in your post as examples.
The first I want to mention is the one I tend to go to first when trying to decide which combination of words may be more common (or whether or not a certain combination can even be found on the web). It is the Google Ngram Viewer. Most are probably already aware of it, but if not, here's what it can show us for "ein Herz aus Gold" vs. "ein Herz von Gold":
Keep in mind that the Google Ngram is limited to searches of four words or less, but if you take time to examine it, it is actually a pretty versatile tool for language study. Note that this example supports what tofro mentioned — that "ein Herz von Gold" would have been perfectly acceptable 150 years ago.
If the Google Ngram doesn't help me and/or I need to do a search of more than four words, I then just do a regular Google search with quotation marks. The number of pages that are returned for each gives me some sense of how frequently each combination of words is used (if at all). Please be aware that just because a combination of words doesn't show up in a Google search does not mean that the phrase is not used or doesn't sound perfectly natural. A random search of some phrases in your own native tongue will be evidence enough of that. Still, this method can be helpful. Again, I'll use the two prepositions you have inquired about as an example:
Also, if such a search still makes me curious, I'll choose a newspaper I respect for its quality of writing (e.g., Der Spiegel) and filter with that or apply a Google Book filter. If I choose to do a Book filter, I look at the names of the authors to try to guess whether or not the book was written in its native language or translated from another. For example, if I were looking up a German phrase, I would want to see lots of books pop up in the results with authors whose names looked German. If I don't see that, it leaves me with suspicions about how natural or common a certain word or phrase may be.
I could also mention the Wortschatz-Portal, which is mentioned in this German Stack Exchange discussion. Though it is extremely interesting, I am unaware of how it could specifically answer your question. If someone disagrees, please feel free to correct me!
Obviously, I have not directly answered your question about what the difference between these two words are. I have chosen not to do so because I don't feel as if I can add to the resources referenced or to what has already been mentioned. However, your post made me think that you (and others) may benefit, both with this specific example and others you may have in the future, from what I've added to this discussion thread. So I hope it helps in some way.