Right, the problem is that by shorting the output the 220uF on the reference drives the reference negative in a big way.
No, 220u isn't normal. Generally I avoid anything above 1u unless it's an actual power application and this issue is one reason it's nice to do that.
First, in general when you're making RC's you can chose the R to keep the C in a 'normal' range. With 10k/1k you have room to go up in R before having problems (>100k or 1Meg typically)
But second I think your AC feedback frequency choice is also off the charts here. I'd change to 220n, or 22n and I expect it will still work.