For better or worse, most commercial embedded development is done in C/C++. If you are already familiar with these, start with Atmel Studio and an AVR 8 bit processor (easier to learn the hardware side, simpler processor) or mbed.org and one of the supported boards (STM32 Nucleos are cheap and have built-in JTAG programmer/debugger with a wide range of processors).
Pay could be higher or lower, you will almost certainly take a cut initially but there are plenty of high paying embedded development jobs.
We have been using Asterisk (SIP/IAX) for about 7 years both on "cloud" servers and field communication devices with no issues on production devices (only minor problems during development/qualification). Our experience has been very positive and the community is very helpful.
If you are looking for speeds on the order of 0.0024 Mbps (2400 baud) then maybe. You could use amateur radio packet radio gateways and radio modems and set something up.
You could do it legally (in the US) by paying a token amount and getting a part 90 business license or possibly by using MURS frequencies instead of FRS/GMRS.
Take a look at Autodesk Fusion 360. It is free to makers / small startups. I use it extensively for product prototype development including for 3d printing and subtractive machining. Another good resource is McMaster Carr, they have downloadable 3d models for most of what they sell. Having accurate 3d models for all the fasteners and other parts makes a big difference in quickly building a 3d model as well as the quality of renderings done in the cad system.
Get a good lawyer, even if you only pay for a couple of hours. There are things you can do immediately (depending on whether you received a letter or they actually filed a suit) that may make them go away.
I was party to one of these a couple of years ago and if we had received good legal advice early it would have been much less expensive in the long run.
As a developer it is worth it as long as you have separate teams responsible for development and production. Segmentation means that a prod server can't accidentally be configured to access a dev resource (db instance, etc). If you have development hand off installation packages and instructions it also means that dev never has to touch prod servers (except read access to logs, etc).
It also eliminates the temptation to "temporarily" use a dev machine/instance as part of prod to solve a problem (and then leaving it that way for years).
I would start with Octave. You can use it to quickly learn the math behind music synthesis and experiment with the various methods. It is open source and available for Windows,Linux,Mac and Android (at least, there are probably other ports).
Then you can concurrently learn a programming language (IMHO based on your post, Python might be a good place to start).
You might also take a look at CSound which is a dedicated environment and programming language for sound synthesis. The CSound documentation pages also have references to a wide variety of other available systems (like Max and Reaktor).
If you are going to ultimately create audio apps for the iphone then I don't think you can avoid C/C++/Objective C but you will have an easier time doing the sound synthesis experimentation in an environment tailored to that function.
It is not a high-end notebook construction-wise, but I have had very good luck with a Toshiba S70 with a full-hd screen. $499 on woot and it has a 17.3" screen, i7, a second sata bay and 2 open DIMM sockets. Battery life is good and it uses a common power adapter. It has reasonable internal speakers (better than average, not a gamer machine). Mint 17 runs like a clock and all hardware is supported by a standard install.
Another non-answer for your question, but I'm not convinced that WiFi is the way to go for IOT. WiFi networks don't deal well with large numbers of clients, power is an issue and setup can be a pain.
POE would be nice for new builds where the cabling is not a problem but I don't think it is practical for anything you would want to retrofit into an existing residential structure.
If I had to bet, 6lowpan is probably going to be the winner in the wireless IOT fabric arena.
What microcontroller/board/etc is best depends totally on your application. Do you need low power, do you need processing power, do you need a significant web management interface?
I have chosen the STM32 ARM line and 6lowpan using atmel zigbee chips but your mileage may vary.
Yes, a significant number of our users still use Windows XP and building things to XP compatibility increases the chances of running under Wine for the 0.5% or so of our users who want to run linux.
Delphi -> C# is a pretty short walk, and there are lots of C#/.NET jobs around. As for learning C#, there are lots of good books and you can download Visual Studio Express and just dive in.
Not with a strong cypher. Any single bit change in the source will result in what looks like a new pseudo-random output.
Determining the private key given the public key in RSA-2048 requires factorizing large blocks of very large numbers which is not practical given current computing hardware. People are working on "smart" attacks that solve the factorization problem from a theoretical math angle instead of a brute force computational approach and there is some consensus that RSA-2048 may be broken in the next few years. A transition to Elliptic Curve Cryptography (ECC) is underway and RSA has never been approved for Suite-A or Suite-B encryption which might lead one to believe that the NSA has known about the weakness of RSA for a long time.