This is related to the network disconnection between Kamioka and Kashiwa on 7th Aug.

Though all raw frames during the network disconnection were sent by the resend process, some automated process missed GWF files at processed time. LL frames during that period are still missing. Note that IFO was out of observing at that time. So there is no urgency to regenerate LL h(t).

Problematic periods were as follows.
- full: [1438571328, 1438572736)
- science: [1438571328, 1438572736)
- second: [1438570800, 1438572600)
- minute: None
- LL: [1438571359, 1438572778)

On the test instance (see also klog#34755), cache files was regenerated properly after the resend process. So I replaced ones in the product instance in which there are some lacks of delayed frames. It seems that replacing the product version of the cache generation code with a new code will not cause any problems.