1. There is no size of data embedding where use of anything other than OP_RETURN or the witness is preferred. Given that most node policies already allow the relay of arbitrarily large data up to the tx standardness size limits (using multiple P2SH outputs), no size limit on OP_RETURN is compatible with the reasonable policy goal to incentivize all data embedding to use OP_RETURN outputs as *opposed* to anything else.

2. Once the OP_RETURN size limit is removed, the restriction on # of OP_RETURN outputs does not accomplish anything.

3. 0.1 sat / vB was chosen among other reasons, since it was below the feerate observed on the network, and above the estimated bandwidth cost an attacker would pay to consume 1 byte of bandwidth from every node on the network, if you'd like to read more: https://delvingbitcoin.org/t/changing-the-minimum-relay-feerate/1886, https://github.com/bitcoin/bitcoin/pull/32959, https://github.com/bitcoin/bitcoin/pull/33106