Azure CPU spike all Sunday afternoon and evening

The Monster Cache server is hosted on Microsoft Windows Azure.  I made that decision for a few reasons:

1. Initially I was just going to use my existing Lanitek server, but I have another important web site running there, GPRA Upload.  Once I had an issue on Monster Cache where I wasn't closing some database connections and it brought down not only Monster Cache, but GPRA Upload too.  That convinced me to get Monster Cache its own server.
2. At the time, I had just recently attended a seminar on Windows Azure, so I guess I was curious to try it out.  And they were offering a 90 day free-trial (or maybe it was 180 day?), so that was good since we were several months away from being ready to launch.
3. I had, and maybe still do have, a hope that if Google or Apple ever promote the game then there will be a spike in traffic, and then I could easily increase the server capabilities on a short-term basis, if needed.

I haven't been super happy with Windows Azure for these reasons:

1. Its more complicated to use than a normal web server.  Some things that took extra time that I can think of off-hand are web Sessions, file uploading, and making backups.
2. They keep changing the features. When I signed up, they had pretty decent tech support.  Now you have to pay extra for tech support, with the minimum plan being an extra $29 / month
3. It hasn't worked perfectly, I've had a couple instances of long down time that I consider their fault.

One instance of down time was yesterday.  I noticed in the afternoon that my Monster Cache app wouldn't log in, but I thought it was my phone's issue.  When I tried the web site at 9 pm that night, I found the web site wasn't even coming up, but just failing to respond.  I started looking for support options, and that's when I found that  I no longer had tech support, just billing support.  So then I looked for a way to "reboot" the app, since there isn't really a lone server to reboot.  That's when I found the CPU monitoring tool.  It showed that Monster Cache had been using 86% CPU since mid-afternoon.  Here's their chart:

That was very strange since I know that our traffic is VERY low, like only dozens to hundreds of hits per day.  There was an option to reboot an instance of a role.  I only have one instance, so I rebooted it at about 10 pm.  You can see that the CPU went back down to 0% and its been working fine so far today.