1. It reminds me the discussion from D2131, where @deadalnix asked for removing the bool parameter because bool parameter is an anti-pattern. I could have factorized things slightly more, but the code being added in the middle of the function it won't gain much.
2. The whole test is assuming that create_test_block won't increment tip height ; as soon as a block is accepted, it is invalidated. Furthermore, the tip height and block time are explicitly set at some points in the test. I think this makes the test flow difficult to figure out, but changing it will require much more changes in the code and is beyond the scope of this diff.

I didn't catch #2 since I haven't finished reviewing this diff. Thanks for explaining.

Regarding #1, that's a good point. In that case, we should be using composition here. Something like to:

def create_test_block_spend_utxos(self, node, txs, version=536870912):
    block = create_test_block(node, txs, version)
    block.vtx.extend([self.spend_tx(node, tx) for tx in txs])
    block.hashMerkleRoot = block.calc_merkle_root()
    block.rehash()
    block.solve()

It's not a massive improvement, but this way, the tipheight and last_block_time strangeness is contained to just create_test_block rather than duplicated in two places.