Josef Widder

> 3\. A QA log with dated entries linking to _manually executed_ testnet/MBT/etc. results. The idea here would _not_ be to capture results from any automated testing, as these are...

> * Is there any concern with initiating a trusted header outside the trusted period (which will most likely be the case)? If the trusted header is outside the trusted...

> For fast sync, where we start our trusted state at genesis, isn't it likely that genesis time will be outside of the trusted period Then we shouldn't do that....

Regarding point 1. is there a way to ask a node for a trusted state? Once the chain is running for more than 3 weeks, we cannot use the genesis...

Shouldn't that behavior be prevented by increasing timeouts (cf. https://github.com/tendermint/tendermint/issues/2920#issuecomment-445874592)?

> * `v_i` and `v_j` have more than 2/3 overlap (less than 1/3 difference) > (this is necessarily true since the update succeeded previously) In principle the reasoning is OK....

> For genesis, I think we can include it, but as "initialisation logic" that needs to be run before any of the handlers or such. I think adding something on...

> This is true, and intentional. Do you think this is a bug, or unclearly documented, or something? At this point it was just surprising, and I wanted to point...

> The client state should store only the absolute minimum information required for verification of subsequent headers We will definitely need `NextValidators` (in addition to the Validators that are already...

I am not sure about the consequences of removing `validatorSet`. Is the idea that once a header is verified we do not need to store the information anymore? Btw. the...