Thread returns on a different wire than it was called on

[ixv]

I have a segment of code that looks like this:

  ?:  ?=([%khan %arow *] sign-arvo)
     ?.  ?=(%.y -.p.sign-arvo)
       %-  (slog tang.p.p.sign-arvo)
       `this
     ::
     ?:  ?=([%volt %get-invoice @ @ ~] wire)
       =/  ship  (slav %p i.t.t.wire)
       =*  tid   i.t.t.t.wire
       =+  !<(upd=update:volt q.p.p.sign-arvo)
       ?>  ?=(%new-invoice -.upd)
       =/  w=^wire  /volt/send-payment/[tid]
       ~&  >>  outgoing-wire+w
       =/  =command:volt  [%send-payment payreq.upd `ship]
       :_  this
       (send-payment:core w command)^~
     ::
     ?:  ?=([%volt %send-payment @ ~] wire)
       =*  tid  i.t.t.wire
       ~&  >>  incoming-wire+wire
       =^  cards  state
         (submit-tool-call:core tid [%pay-success ~])
       [cards this]

ie: you call get-invoice, it returns and then calls send-payment. But the wire that you use when calling send-payment is not the wire that send-payment returns on.

From the dojo:

>>  [%outgoing-wire /volt/send-payment/thread_zoUyuybqWKx6UXSLziRymMn2]
>>  [%incoming-wire i=%~.volt t=[i=%~.send-payment t=[i=~.thread_yQg1uWvMycSezjVH1tFD2bw8 t=~]]]
spider got input for non-existent ~[~.khan-fyrd--0v2.bp61q.rns9h.b86rq.o79lv.8qvfk]

That second wire doesn't come from nowhere though, if I do this twice in a row, the incoming wire will match the outgoing wire of the previous run.

But if I nuke spider and revive it, it hangs after the send-payment call, printing the outgoing wire but not the incoming one.

Seems like maybe there's some stale state in spider, but I don't know how to reproduce this from scratch. If someone wants to look into this I'm down to jump in a call and give you access to the ship its running on.

[pkova]

If you chop and zip up the pier we'll take a look. DM or email me with the info.