For security, we are planning to load all our textures from a secure CDN using temporary URLs. These URLs will expire after15 min typically. From what I can tell, Babylon caches textures in memory using the url as the cache key. Since the url for the same asset could potentially be different at different times, using url as a cache key would result in more downloads than necessary. Our plan is to implement our own cache that will cache a Texture instance by our uid for the texture and pass that when we need to assign it directly (e.g. to a material). Do you see any issues with this approach? Is there anywhere Babylon depends on having the texture cached in memory by url?
Only place is the cache actually so it should be good 