MWEB is certainly an improvement over transparent transactions (and other methods such as coinjoin, coinswap, cashfusion, etc.), and I welcome the litecoin upgrade. I agree that decoy-based privacy is weak.
However, I don't believe that the mimblewimble meets the standard of privacy needed for most users. It's not the visible amounts and addresses, but the links between transactions that are the main problem. CTs on their own are just a "nice-to-have".
The end goal should be a zcash or firo style of privacy. I think you can scale that to a global network with an adjustable block size, payment channels, and atomic swaps between multiple cryptocurrencies. The problem is that zcash and firo have weak tokenomics compared to monero. Grin will have a hard time finding an initial niche that isn't currently satisfied by monero, and if it does take off, its changes could be merged into bitcoin (https://www.truthcoin.info/blog/imex/).
Everything considered, I don't think that the risk of a monero inflation bug is greater than a bitcoin inflation bug when you consider the complexity associated with scripting.
* node resources scale with the size of the UTXO set (unspent outputs), which in Monero's case balloons to the entire TXO set (all outputs, orders of magnitude larger)
* a typical 2-input 2-output transaction is 4 times larger
* wallets have to track all outputs to choose random decoys for transaction inputs
One can argue that this is the price to pay for significantly better privacy, but the largest benefits come from having no visible amounts or addresses, which can be achieved with significantly better scalability than BTC [1].
[1] https://forum.grin.mw/t/scalability-vs-privacy-chart/8114