Fix usage of threads for pages migration
What does this MR do?
The main thread already is one that is running
.
The child threads should always wrap only operation
instead of the whole thread. This ensures that wrapping
operation can hold a lock when loading code.
In a worst case scenario it will result that initially till all threads do load all relevant code it will be run effectively in sequence. This will resolve once all dependencies get loaded lazily.
The failing and correct example can be found here: #300123 (comment 500807759).
Does this MR meet the acceptance criteria?
Conformity
Edited by Grzegorz Bizon