Fixes an assert when we start building a trace but then have to abort due to the private copy failing.
There is still a long-term problem here with the private copy failure potentially happening over and over: that is covered by #4038 (closed).
Tested on the linux.rseq tests with new (forthcoming) instrumentation for #4019 (closed) which reliably hit this assert.
Issue: #2989 (closed), #4019 (closed), #4038 (closed)
Fixes #2989 (closed)