ID 567762 - From Editorial: Publishes back are 30x slower than they need to be due to lack of parallelisation

Follow

Problem summary
For a large amount of the work performed during a publish back to Flix, only one CPU core is being utilized, which results in imports being much slower than they need to be. For example, if the CPU has 30 cores, a 30-minute long import could take 1 minute (in the ideal case).
 
Customer reported version
flix_6.6.2
 
Customer reported platform
windows10
 
Steps to reproduce
  1. Start a large publish back to Flix from editorial.
  2. Attach a profiler to measure how well Flix is utilizing the CPU cores.
 
Expected behaviour
Flix makes good use of all available CPUs.

Actual behaviour
Flix does most work on a single core 90% of the time.
 
Workaround
None, Flix can't split the AAF into small parts and import them all at the same time since simultaneous imports can cause deadlocking issues in the panel and revision counting code.
 
Reproduced by support
Flix 6.6.2
Flix 7.0.0

    We're sorry to hear that

    Please tell us why