What do you mean by this?  Do you mean that you found two 256-bit ECDSA public keys that hash to the same 160-bit address?

How is it relevent that the process was random?  To show a collision, it is necessary and sufficient to show two preimages that hash to the same value.

In what sense do you think that Bitcoin is non-fungible?  Merriam-Webster defines "fungible" as "being something (such as money or a commodity) of such a nature that one part or quantity may be replaced by another equal part or quantity in paying a debt or settling an account".  Bitcoin seems to meet that definition; basically all entities who accept bitcoin will accept payment from any UTXO.

Most software is crap and easily hacked.  But it's definitely possible to create secure software.  Some software, such as the seL4 kernel, is even formally verified.  Even simply running a mainstream static analysis tool on your code and fixing all the warnings (which of course is easier said than done) is enough to remove most opportunities for hacking.

Because there is no way to do this trustlessly, and the counterparty risk is too much (not just in terms of trusting the counterparty to be honest, but trusting their cybersecurity as well).

This is where I stopped reading.  (ಠ_ಠ)

Changing the block-generation reward to allow more than an asymptotic limit of 21 million coins would require a hard fork.  It would be a contentious hard fork, because many people strongly believe that the original plan should not be changed to dilute existing coins.

Then the smallest unit of bitcoin (1 satoshi) would be worth $0.01.  It is possible to hard-fork Bitcoin to allow further subdivisions; this would be far less controversial than increasing the maximum coin supply.

Bitcoin Core has supported SegWit since version 0.13.1.

There's some speculation that people want to get bitcoin before the fork so that they get money on both the incumbent chain and the SegWit2x chain.  After the fork, it's to be expected that the price will drop.

A single word randomly chosen from a list of 2048 words provides 11 bits of entropy (2^11 = 2048), so 24 words provides 264 bits of entropy.  These 264 bits can be used as the seed for a cryptographically secure pseudo-random number generator (CSPRNG).  The CSPRNG can then be used to produce an arbitrary amount of public/private key pairs.  A good CSPRNG has the property that it is computationally infeasible, given only a subset of the outputs of the CSPRNG, to determine the state of the CSPRNG or other outputs of the CSPRNG.  This is how it is possible to "stretch" the original 24 words to produce an arbitrary number of addresses.

Most ICOs already need to be registered with the SEC under existing law, because they constitute securities under the Howey test.  The SEC has recently launched a cyber unit and has already begun enforcement actions against some of the most prosecutable targets (e.g., blatant scams with known principals).

https://www.sec.gov/news/press-release/2017-176
https://www.sec.gov/news/press-release/2017-185-0

Google Translate for the relevant paragraph is:
"b) the establishment of requirements for the organization and implementation of production based on the principles of cryptography in the medium of distributed registries ("mining"), including the registration of economic entities carrying out such activities, and determining the procedure for its taxation;"

I'm not sure if "economic entities" ("хозяйствующих субъектов") includes private individuals, but my wild guess would be no. 

(The original Russian text of the above-quoted paragraph is:
"б) установление требований к организации и осуществлению производства, основанного на принципах криптографии в среде распределённых реестров («майнинг»), включая регистрацию хозяйствующих субъектов, осуществляющих такую деятельность, а также определение порядка её налогообложения;")

Bitcoin Gold supposedly has replay protection, so at least in theory, after 6 confirmation of block 491407, you should be able to spend Bitcoin from a Bitcoin address while keeping the Bitcoin Gold on that address.

The textbook for that course is also available online:
https://freedom-to-tinker.com/2016/02/09/the-princeton-bitcoin-textbook-is-now-freely-available/

Another suggestion is to try the homeworks and exams from the Stanford course: https://crypto.stanford.edu/cs251/

I don't understand what threat model this is protecting against; can you please elaborate?  If your OS is compromised with malware that can steal your encrypted wallet and your password, can't it just as easily steal your plaintext wallet after you decrypt it?  Although hardware keyloggers that sit between the keyboard and the computer do exist, this requires an attacker to have physical access to your machine, and in that case you're basically screwed anyway.

A fresh install of MS Windows is certainly more secure than an OEM install that is potentially compromised with crapware such as Superfish.  But Windows itself still has a huge attack surface.  Personally, I'd recommend a smaller, security-focused OS such as FreeBSD.

Most ISOs have signed hashes that you can verify after downloading (e.g., https://www.freebsd.org/releases/11.1R/CHECKSUM.SHA256-FreeBSD-11.1-RELEASE-amd64.asc).  Of course, this requires you to trust the public key that was used to sign the hash, but the chain of trust must be rooted somewhere.

IANAL, but it seem to me that each payment would be taxed as ordinary income, provided that the received asset is liquid and there is a well-functioning exchange to determine its fair-market value (in USD).  The IRS FAQ (https://www.irs.gov/pub/irs-drop/n-14-21.pdf) mentions that it might also be subject to self-employment tax if carried on as a business as opposed to a hobby.

Best bet is to use a clean machine with a wired keyboard.  (Wireless/Bluetooth keyboards have a greater attack surface.)  A simple trick is to not enter your password completely sequentially, but instead to enter some of it, then reposition your cursor using the mouse, and then type the rest of it.  This will likely thwart attacks not targeted specifically at you, at least as long as this technique isn't widely used. 

To brute-force pre-image resistance of hash functions such as SHA256, a quantum computer would still need to perform a number of operations proportional to the square root of the number of operations performed by a normal computer.  Therefore, simply doubling the number of bits (e.g., switching from SHA256 to SHA512) would provide quantum-resistant security equivalent to existing security against normal computers.  (Note that sqrt(2^512) = 2^256.)

The vast majority of Bitcoin UTXOs include only a hash of a ECDSA public key, not the ECDSA key itself.  Therefore, if the public key has not already been revealed, an attacker would need to (1) sniff a transaction as it is entered into the mempool, (2) crack the private key, (3) create a new transaction using the private key, and (4) get this new transaction committed to a block before the legitimate owner's transaction.  To guard against this attack, Bitcoin would need to add support for quantum-resistant asymmetric cryptography.  Unfortunately, this field is still immature, and existing quantum-resistant asymmetric cryptographic schemes (see, e.g., https://en.wikipedia.org/wiki/Post-quantum_cryptography) are much less practical than ECDSA.  But it is likely that they will develop to be the point of being practical before quantum computing poses a severe risk to Bitcoin.  In fact, it is not even known for sure that quantum computing will ever economically scale to the point where it can attack 256-bit ECDSA.

I have doubts that it would succeed at the above goal (decentralization of mining) if it is successful in getting significant market capitalization.  Once it becomes very profitable to mine, people who live in areas with cheap electricity will heavily mine it to the extent that it would be unprofitable to mine for people with average electricity costs.