>From what I can tell this is the same issue Thomas was looking at yes. I saw the failure on the master branch at the v9.2.0 tag (ae35f033) and just re-tested it against (83aaec1d) and still see it. I haven't seen it be an intermittent failure, it has failed 100% of the time that I have tested it when targeting arm-softmmu.
However, you are right that with Phillippe's series on top of my change the bug is somehow re-introduced. Not sure what the protocol is here since neither patch has landed