You are right these are heuristics. However they may be amended with data from other sources. 

Wasabi was great to address issues of WabiSabi in multiple versions.

ps: @kruw you might want to look at [this](https://github.com/crocs-muni/blocksci/blob/d30dd09c8908fe0bf27fe9115bac1917d820d6c8/blockscipy/src/chain/coinjoin_module_py.cpp#L226) lol

That's just the heuristics of detecting whether a tx is a wasabi2 coinjoin. There's also a `isLongDormantInRemixes` above that that attempts to identify based on the time between remixes whether the cj is initiated from trezor suite.

There are some deeper heuristics defined [here](https://github.com/crocs-muni/blocksci/blob/d30dd09c8908fe0bf27fe9115bac1917d820d6c8/blockscipy/src/chain/coinjoin_module_py.cpp) which are about as far as you'd get. But are these weaknesses in the wasabi protocol? Or are they patterns that can be broken by behavior?

optimism

privacy

Privacy in Bitcoin, state of the practice

satsbridge

A set of CJ-related heuristics implementations is here. 

https://github.com/crocs-muni/blocksci/blob/d30dd09c8908fe0bf27fe9115bac1917d820d6c8/src/heuristics/tx_identification.cpp#L601