Imagine an application that updates an array rarely and looks up often. Say you have 10,000,000 customers / books / whatever that get added / published only once a minute, but information is requested 1000 per second. So, re-building the hash for lookup 1000 times a second is clearly not an option. Right?

Um, if the application is as you describe, why would you want to use an array at all? With that sort of ratio between updates and searches, it would be better just to maintain a hash instead of an array. Surely you do not want to "recreate the hash every time" in order to implement the search; create the hash once and maintain it, assuming it fits in memory -- and if not, consider a DBM_File approach (cf. AnyDBM_File), or just use a real database backend.

If your issue is "using a (temporary) hash to store a copy of an array so that the array can be searched for specific values", yes, that's a bad approach for really big arrays and the kind of update/search ratio you're suggesting. But you're not explaining why the primary data storage needs to be an array.

If the issue is really "coming up with a viable app to support lots of searches for specific values in a large set", the answer is more likely to be: start by using a hash as the primary storage, and use the values to be searched for as hash keys.