"Sufficient Reliability" still needs mechanisms to resolve the rare issues that occur. A 5 9's system still needs resolutions for the 0.001% scenarios. In the case of your phone's alarm if your phone crashes, runs out of battery, app cache gets corrupted, whatever-- you can intervene with your human judgement to determine an appropriate resolution. What mechanism do smart contracts have for 0.001%?

Your phone's alarm is also not a critical transaction worth a person's life savings or the wealth of a country or a transaction involving life-critical supplies etc. The bar is a bit higher here, and even the most reliable systems ever designed have the ability to insert human judgement when the rare issue happens. Smart contracts are supposed to be appealing because they avoid the need for biased/imperfect judgement in favor of something "trustless", or at least that's the vision many see for them.

Contracts can also have hundreds or thousands of clauses, making "sufficient reliability" a much higher bar than an alarm clock. Especially because many of those clauses entail human concepts that would be extremely difficult to translate into code: What is the algorithm for determining "force majeur"? That's a pretty basic clause that appears in many contracts, but I don't know where you'd even begin to get a computer to understand & properly identify such events.

I don't see a pathway to sufficient reliability in smart contracts anywhere on the horizon, save for very simple cases. Even then, here we have IRON, which should have been relatively simple as these things go, but failed because the simple case of "Titan has no value" was not considered.