I would not like to see a general purpose OS adopting a thread-per-core model, so you are hardly wrong.

However, applications are, by design, not general purpose. If you are building an application that serves a specific purpose the model may fit you well.

Some models, like TPC, are niche. But as changes happen in the industry and technology the niche can become smaller or larger.

In the case of TPC, that niche is now larger for two main reasons: As processors go multi-core, the cost of communication between entities not sharing the L1 increase (technology). And as applications working through the network become the norm, scale out architectures that shard their data are more likely to be adopted (industry). Such applications lend themselves very well to a thread-per-core paradigm