Hi Kiart, welcome 👋
• the self-hosted open source version is essentially the Starter version, though without the seat limits. You can upgrade the self-hosted versions to the 2 paid plan levels, Pro and Enterprise, with a license key
• Yes, there are actually a few ways to do this. We have a notion of ‘phases’ which can be used to date limit experiments. You can also do ad-hoc reports to adjust the dates
• Yes, Bayesian is the default engine (and we have Frequentist and Sequential too)
• Performance is a bit hard to judge - I’ve seen good results from both. I have a gut feeling that Redshift might be better, but there are so many variables and ways to optimize data with partitions and indexes, this usually is the larger factor for optimizations.