We have a special initialization procedure compared to other miners which is more sensitive to configured clocks etc. For pushed mem clocks, we sometimes get these retries before succeeding. You want those clocks for your mining though, so we had to add logic for handling failed inits with automatic retries. If mining is working fine after that, you're all good, although I know it might look a little weird with these error messages and retries.

Maybe I'm misinterpreting things though - are you saying it takes 2-3-4 attempts to get a "successful" start like the above, and you have to restart a nr of times manually to get there? Or, is it starting 100% of the time in the end, but with the errors and retry attempts in the logs above?