Comment by film42

Nice! In rails I end up using Rails.cache most of the time because it's always "right there" but I like how you break out the cache to be a per-method to minimize contention. Depending on your workload it might make sense to use a ReadWrite lock instead of a Monitor.

Only suggestion is to not wrap the error of the caller in your memo wrapper.

> raise MemoTTL::Error, "Failed to execute memoized method '#{method_name}': #{e.message}"

It doesn't look like you need to catch this for any operational or state tracking reason so IMO you should not catch and wrap. When errors are wrapped with a string like this (and caught/ re-raised) you lose the original stacktrace which make debugging challenging. Especially when your error is like, "pg condition failed for select" and you can't see where it failed in the driver.