Another consideration that seems to have been obscured by the debate on privacy and the narrow focus on the particular client implementation of the app is the significant problem of false positives and negatives.
A lot of voices have spoken out about this issue overseas (particularly in the US) while many local tech voices have skipped considering this at all.
We need to be driving test numbers up until only 3-5% are showing positive in order to be confident about low prevalence in an area. I don't see a problem with false positives encouraging asymptomatic people to get tested - it's as good of a sub-population as any.
The project lead of Singapore's TraceTogether initiative goes into detail about the problems with an automated system, and why a human-in-the-loop is ideally required to evaluate the type of contact and make a determination.
A determination around being a close contact results in 14-day isolation regardless of symptoms, presumably because you may initially test negative before moving into an infectious and asymptomatic or symptomatic phase.
I assume that the use model is as a way to help human contact tracing, not to replace. Admittedly, the benefit is less then, but maybe it's still worth it?
The assumptions of benefit appear to rely on a naive theory of instantaneous contact and isolation, when in practice the entire process is unavoidably manual, requiring human intelligence to ascertain environmental factors with which to make a determination about whether someone is a close contact.
When you consider that close contacts include anyone you've spent more than 2 hours with in a room, it becomes clear that most life situations are not handled by estimates of proximity using bluetooth: home, family & friend visits, workplaces.