There are two known issues with the checkers. Or actually more with the infrastructure behind them.
1) Sometimes the op-cache gets corrupted. We are clearing it once per hour as a counter measure. Until a better solution has been found that is.
2) Sometimes the service just stops responding. We are a bit more clueless here, but we do restart it once per day. This seems to be a very rare issue though.
