Riot/Nightly spontaneously cancelled cross-user verification from Riot 1.5.14

[ara4n]

@matthew:arasphere.net on Riot 1.5.14 tried to verify @matthew:matrix.org on Nightly, and got cancelled. The verif never appeared as a toast on Nightly.

The first event was:

{
  "content": {
    "body": "@matthew:arasphere.net is requesting to verify your key, but your client does not support in-chat key verification.  You will need to use legacy key verification to verify keys.",
    "from_device": "DKAZBAEEPK",
    "methods": [
      "m.sas.v1",
      "m.qr_code.show.v1",
      "m.reciprocate.v1"
    ],
    "msgtype": "m.key.verification.request",
    "to": "@matthew:matrix.org"
  },
  "event_id": "$15856596317WcHyY:arasphere.net",
  "origin_server_ts": 1585659631882,
  "sender": "@matthew:arasphere.net",
  "type": "m.room.message",
  "unsigned": {
    "age": 9138
  },
  "room_id": "!PcWroKvERfdTcjBJCK:matrix.org"
}

[bwindels]

The client for @matthew:matrix.org (the receiver) seems to get confused about the .request event after changing the phase:

2020-03-31T13:00:41.164Z I Verification request $15856596317WcHyY:arasphere.net: m.key.verification.request event with id:$15856596317WcHyY:arasphere.net, content:{"body":"@matthew:arasphere.net is requesting to verify your key, but your client does not support in-chat key verification.  You will need to use legacy key verification to verify keys.","from_device":"DKAZBAEEPK","methods":["m.sas.v1","m.qr_code.show.v1","m.reciprocate.v1"],"msgtype":"m.key.verification.request","to":"@matthew:matrix.org"} deviceId:undefined, sender:@matthew:arasphere.net, isSentByUs:false, isLiveEvent:true, isRemoteEcho:false, phase:1=>2, observeOnly:false=>false
2020-03-31T13:00:41.165Z I Got sound default for !PcWroKvERfdTcjBJCK:matrix.org
2020-03-31T13:00:41.180Z W Cancelling, unexpected m.key.verification.request verification event from @matthew:arasphere.net
2020-03-31T13:00:41.181Z I sendEvent of type m.key.verification.cancel in !PcWroKvERfdTcjBJCK:matrix.org with txnId m1585659641181.220

[bwindels]

It looks like we're processing the same .request event twice for some reason ... perhaps a race with decrypting the event?

[bwindels]

Will have a further look after finishing the QR stuff.