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.
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.