I can tell you exactly what the problem is, MTGO is a legacy product.
You see, to corporate executives, any video game older than 18-24 months is a "legacy product" and is not worth spending more than a pittance on, period. This is why MTGO was still using the 1998 interface and had all the crash problems, it was designed for a much smaller user base and they simply never got budgeted to expand it.
Er, no. That logic would make sense if they didn't continuously port every single new card printed into the MTGO rules engine (1000 cards a year or more), design the specific wordings of paper cards around the online UI, etc. This is also the 4th client release, with the client being built from the ground up each time.
The problem is that they have a policy of not fixing things until they're REALLY REALLY BROKEN. This isn't just MTGO specific, it goes for just about every organized play policy in the history of the game, and elements of card design as well.
The problem with the servers is that they never invested in any kind of load balancing infrastructure. They just threw more/newer hardware at the problem, but Moore's Law in the simplest sense of faster processors didn't keep up with the growth of the game or their ambition to host really large events online. It's a simple load scaling problem, and I'm willing to bet their server code for event hosting isn't even multithreaded, let alone takes advantage of virtualization to spawn virtual servers based on demand.
However, there's a hidden level of truth to what you were saying. Hasbro treated all of MTG, paper and digital, as a legacy product for a long time, content to milk some money out of the remaining players by pushing out a few sets a year. Then, about 4 years ago, it started roaring back to life, and then soon after, started making a lot of fucking money. WOTC is Hasbro's most profitable division currently, while tradition toys have taken a nosedive for them.