Tough question. There are several "features" of bitcoin that seemed good ideas at the time; but now, in hindsight, they seem to be mistakes, and Satoshi himself would probably agree.

One of them is the expectation of huge increases in value, resulting from the fixed cap and the assumption that it would replace other means of payment.  That turned bitcoin into an allegedly safe and lucrative investment, and a speculation tool.  Most of the big problems that bitcoin is facing today are a consequence of that undue retargeting of the project.

So, one problem that would have to be fixed is making it unappealing to hoarders and speculators.  

Then the value of 1 coin would be determined by its use for e-payments, according to the money velocity equation.

Another problem is ensuring that its value stays in a bounded range, so that ordinary payments can be expressed without too many zeros before or after the point.  Say, between 0.01 and 10 USD.  In the first extreme case (about the value of 1 yen) one would not need decimal fractions, and do all accounting with integers. In the second extreme case, one could truncate after 3 decimals (i.e. use the "milly" as Americans use the penny).

A stable value would seem to attract hoarders and investors, but hopefully the two goals can be achieved with built-in demurrage (negative interest).  Namely, each UTXO loses value at (say) 2% per year, so you can spend only 0.98 BTC of that 1 BTC that you earned 1 year ago.  Those 0.02 BTC would implicitly go to the "Bitcoin Treasury" and would be redistributed as block rewards (so the block reward could remain constant forever while the total amount of coins in circulation would still be bounded.)  

Hopefully that negative interest would be enough to dissuade hoarders and speculators.  

However, in order to keep the value stable, the issuance (e.g. by block rewards) would have to be adjusted to the volume of payments and to the mean time between payments with the same coin. I have no idea how that could be done in a way that would resist manipulation by hackers (computational or financial).  It seems that another Satoshi would have to come out with another ingenious invention...

Another problem is the centralization of mining.  That could be fixed by keeping the block rewards too small to make mining into a profitable industrial activity.  But that may conflict with the need to put more coins in circulation as adoption grows.

Other problems are the centralization of all transactions of the world in one blockchain.  That design limits the scalability of the network and makes nodes expensive to operate.  That could be fixed perhaps by having a large number of separate blockchains, each containing the transactions of a certain subset of (say) 1000 addresses.  Then a 1-in, 1-out transaction would have to be recorded in only two of these small blockchains.

Other technical problems include rewarding all players (not just miners), ensuring that every transaction gets processed, replacing the UTXO-based accounting to the balance-based accounting that banks use, etc.

There are also non-technical problems like anonymity, non-reversibility, lost coins, legal jurisdiction, ... but each would be a long discussion in itself...