Bitcoin is already divisible to 8 units, and that's more than enough for the world's population. 1 satoshi would cost around $1 when BTC hits the millions, anyways. That's still cheap when you compare it to a wire transfer or even credit/debit card fees. We don't need to worry about this, especially when it's going to take decades before BTC goes well above $1m per coin.

If there's ever a need to make BTC more divisible, it will come to a consensus among developers, miners, and node operators alike. A hard fork would likely happen by that time, but it won't be as controversial as the big blocks debate by Bitcoin Cash proponents. Everything will go smoothly without issues (at least that's what I hope for). Who knows what would be of Bitcoin in the future? Just my thoughts

As to why the OP believes we have to hard fork the bitcoin project in order to have smaller units, I don't understand the reasoning behind it. When Bitcoin's price rises up when it crosses one million dollars, then it is still appropriate to use a satoshi as a unit of measurement when talking about bitcoin amounts, and it is important to note that doing a hard fork does not have anything to do with having other bitcoin measuring units because it is simply about placing decimal places rather than doing a hard fork.

its not just needing a hard fork to add a new tx format that measures in microsats.
you would also need ALOT more code and rules that treat two states of transactions different.

pre-fork dated utxo treated as sats and post fork tx treated microsats.. separating the classes..  and a whole lot of code to ensure the spending of old utxo's dont ever generate more units when being spent or the opposite

Eg a legacy utxo from 2020. of 10units(sats) where by the new tx format doesnt accidentality end up getting 10 units(10 microsats) and instead would need to be a future utxo of 10,000 units(microsats)

where by if the new tx format then wanted to spend its 10,000microsats units to a legacy(8dec) address ensuring that the legacy address validation code does not treat it as 10,000sats when it should be 10 sats

after all we cant really go back and change all blockdata to be 1000x units compared to norm. so there would need to be alot of CLUDGY code. that comes with alot of risks of generating new units outside of the normal coin reward mechanism

in short. we should not even attempt it. as it is not needed anyway

LN is not the solution.. its too buggy even after 5 years

far easier to start from scratch with a different subnetwork designed differently to LN without the bugs or payment model of LN.. but offers a more secure method of pegging to a 1:1000 unit level to use on the sub network

whilst also reminding people to be risk aware of any sub networks and not pretend a subnetwork of any form is the same as the mainnet

If LN is the solution for this kind of issue, I am curious if LN indeed accept a fraction of satoshi while the original network rejects such kind of transaction, how can that transaction be presented on Bitcoin network since I believe LN transactions need to be reflected on the original network?  Please kindly enlighten me since I am really confused about this kind of situation. 

Good question.

So the transactions involving millisats aren't broadcast to the main chain. They remain entirely within the Lightning network. Whenever a Lightning channel which has a balance involving millisats is closed, then the transaction which is broadcast to the main chain is rounded down to the nearest whole satoshi.

Think of it like this. You and I each put ten $1 bills in a jar. There are now twenty $1 bills in that jar. You sell me some goods for $5, and so we agree that when we open the jar again, you'll get fifteen of those bills, and I'll get five. All good so far. Then I sell you some goods for $2.50. So now the balance is $12.50 for you, and $7.50 for me. If we open the jar now, then we can't get the exact amounts we are owed, because there are only $1 bills in the jar and no quarters, dimes, etc., so we can't receive 50 cents each. But no matter, we aren't opening the jar yet, so it doesn't matter for now. We can agree that this is what we are each owed, without actually opening the jar. So we do another trade for $2.80, and now the split is $15.30 and $4.70. And another trade, and another, and another. All the time we agree to split the original $20 in different ways, some of which won't actually be possible because we don't have the coins needed to split up a dollar in to smaller parts. When we finally come to open the jar, let's say (for example) the split is $15.30 and $4.70, as above. You would get $15, and I would get $4, with $1 left over in the jar.

The same thing happens with Lightning channels. You open the channel using a whole number of sats, but then can trade in millisats. As long as your channel stays open (i.e. we don't open the jar), you can trade back and forth in millisats as long as you like. When you want to close the channel (i.e. open the jar), then the transaction broadcast to the main chain will only use sats, and so the balance of both parties will be rounded down to the nearest whole satoshi, with the extra satoshi (the $1 left over in the example above) added to the fee of the closing transaction.

That will never happen, there is LN. One solution for many problems, though it's not decentralized, it's more like visa and paypal but with less fees.

Here's a quote from the comments in the very code that you just linked to:

Which is exactly what I said. If you close a channel which is holding some number of millisats, then the transaction broadcast to the main chain will be rounded down to the nearest whole sat. Or here is another quote from BOLT #3:

So no, what I said is very much not wrong.

There has never been in the history of bitcoin that a hard fork bitcoin has ever become successful, the ones they did in the early days of bitcoin and the ones they did in the all-time high price of bitcoin to date have not reached the half potential of bitcoin talk more of doing what they are been created for. A hard fork of bitcoin is not necessary nor indispensable, it will make some bitcoin holders if they airdrop it for them but it will mostly enrich some Cartels who will create such fork coins, we have seen it in the past and they didn't it age well. Look at BitcoinCash creator, Craig Wright, he ended up scamming people with his fake Bitcoin only to come back again to say that he is the real Satoshi, a very dishonest person.