Comment 13 for bug 1735154

Finally!

After matching against working cases I could see that when the test fails, it's due to some commands of the transaction don't get commited and some others (the first(s) ones) getting commited in what is supposed to be a different transaction (actually different test running in the same worker thread). Turns out to be that it's a ovsdbapp bug:

1. T1 creates a new transaction txn1
2. T1 adds a new command to it C1.1
3. T1 yields the execution (due to some I/O operation for example)
4. T2 requests a transaction so txn1 is returned as a nested transaction
5. T2 adds a new command to it C2.1
6. T2 adds a new command to it C2.2
7. T2 commits the transaction
8. T1 adds a new command to txn1 C1.2
9. T1 adds a new command to txn1 C1.3

Reported the bug there too:
https://bugs.launchpad.net/ovsdbapp/+bug/1793499

\o/