The standard answer is that everything you listed will restart the Learning process: anything that impacts delivery or targeting will have an impact on performance and thus require FB to re-establish its algorithm coefficients for the campaign.
I generally try to avoid making any changes during the learning period except to simply kill a campaign that's obviously not working. The problem here is that you can only really test one approach / strategy at a time if you're targeting a specific audience, as you wouldn't want to spin up multiple campaigns with different parameters aimed at the same narrow audience. That said, you only need 50 conversions to establish the campaign coefficients, so it shouldn't take too long to exit the learning phase for any given campaign.
So to specifically answer the question, what I understand to be changes that will cause a campaign to restart its learning phase:
- Changing the bid strategy (target cost, bid cap, cost cap, lowest cost, for VO: min-ROAS or maximum value);
- Changing the bid strategy cost parameter (eg. bid cap amount);
- Changing any targeting parameters (eg. geo, phone OS, etc.);
- Changing any audience parameters (eg. changing the Custom Audience);
- Changing the campaign budget (for CBO campaigns) or ad set budget;
- Changing the AEO event;
- Adding / Removing a creative;
Of the settings you listed that may cause the campaign to re-enter the learning phase, what I've heard is that a change of more than 10% will reset the campaign's coefficients and cause it to start exploring again even if it doesn't explicitly re-enter the "learning phase."