BU facilitates breaking consensus, not enforcing it. That's the whole point of BU.

No-one will be able to trust the validity of on-chain transactions, as BU can fork or reorg itself unexpectedly at any time. The ability to alter the consensus rules on the fly turns forks and re-orgs into a feature, when they're highly undesirable for a stable uniform blockchain.

Your point about consensus depends on one's point-of-view.

To quote BU:

We believe that the block size is NOT part of blockchain consensus...
By moving the blocksize limit from the protocol layer to the transport layer,
Bitcoin Unlimited removes the only point of central control in the Bitcoin economy -
the blocksize limit - and returns it to the nodes and the miners.

You are certainly correct that it is not tested. 

I suppose that supporters of BU like myself feel that the
risk is small compared with the damage that is likely to
come by allowing the current impasse and stagnation to continue, and/or allowing core to continue its governance.

Then they're out of their tiny little minds.

From the Bitcoin source code, consensus.h:

Me?  No that is a quote from bitcoinunlimited.info.

Obviously the limit is in the codebase, and we think removing it is a good idea.
So from our point of view, this facilitates consensus, not breaks it.

Perhaps you still haven't gained the ability to discern facts from opinions.

50 megabytes of mempool constantly since a week on my full node. I think that the only decent way to trim well this huge amount of unconfirmed transactions will be to raise the minimum relay fee to 0.0001 BTC per kB