Bob Jenkins has a few varieties of hashes which are pretty fast, and might be a good replacement for $hash().
Thanks for the suggestion. I performed some tests to check hashing algorithm speed/distribution, with different lengths/types of data, eg. short/long strings, numbers, different table sizes, etc. The current hashing algorithm distributes okay in some situations but is awful in others. I tested different algorithms, eg. lookup3, DJB2, FarmHash, xxHash, MurmurHash, FNV-1a, etc. and have settled on FNV-1a for simplicity. In all of the above tests, it worked just as well as the more complex algorithms and had good distribution. It is difficult to say how this change will affect hash table performance in general - most overhead is due to the scripting language and other features. But in specific contexts and with certain types of data, it could make a big difference. This will be in the next beta.