Summarizing what @Raghavendra-PegaSys explained, our current model is (for the sake of PoC simplicity) akin to a traditional vanilla mutex: the owner of the mutex is blocked from acquiring the mutex again.
But we expect it would be easy to generalize to the case of recursive mutexes, which would allow any order of transactions, since any transaction would be able to reacquire locks transparently. This removes yet another burden from the application design.
2 Likes