spun out of this comment

if possible, before finishing the tastora tests, we should check if all validators produced blocks for each version that validator used in the test, and we also need to check if any node in the network halted. If either of these things are false, then the test should fail.

Note: Depending on the number of block producers, there could be some flakiness if not enough blocks are produced per version. Therefore, the number of block the network produces at each version should be at least 3x that of the number of potential block producers. Assuming even voting power, we should then expect each validator to have at least 3 chances to produce a block.