The issue was observed while working on different issues. Reducing the time-trigger delays on Global L0 makes producing global snapshots much faster than metagraph enqueues binaries. Such a situation can trigger the "Retry mode" when the number of pending binaries on Metagraph is lower than the initial cap.
It causes the "Retry mode" to fluctuate so it will constantly enter and exit the retry mode and reset the cap to 3 over and over. The additional condition for exiting the "Retry mode" has been introduced in that PR and unit tests are adjusted accordingly
The issue was observed while working on different issues. Reducing the time-trigger delays on Global L0 makes producing global snapshots much faster than metagraph enqueues binaries. Such a situation can trigger the "Retry mode" when the number of pending binaries on Metagraph is lower than the initial
cap.It causes the "Retry mode" to fluctuate so it will constantly enter and exit the retry mode and reset the cap to 3 over and over. The additional condition for exiting the "Retry mode" has been introduced in that PR and unit tests are adjusted accordingly