Because of multiple human errors, we couldn't back to OBSERVING for a while.
To prevent repeated same mistakes, I summarized what was bad and what must be improved.
Because partial commissioning will be resumed, it'll become a good lesson for many people who will commit commissioning activities.

Related klogs
- Trigger work: klog#34520
- Calibration works: klog#34530, klog#34533
- Initial notice: klog#34536
- Clearing work: klog#34537
- Confirmation: klog#34538
- Back to Obs.: klog#34540

points to be improved
1) SDF clearing work was done NOT in OBSERVATION on LSC_LOCK guardian.
=> We must ensure that all set point values in observation.snap are accepted as ones in OBSERVATION state.

2) A procedure of loading down.snap/observation.snap cycle was skipped in the clearing work.
=> A loading cycle is necessary to mitigate accidental SDF issues.
Though a main purpose of this procedure is to avoid the SDF issues due to numerical rounding errors, it can also prevent a mistakes such as above.

3) A calibration measurement was started with insufficient checks.
=> Calibration measurement must be started after confirming that CALIB_NOT_READY or READY on IFO guardian is reachable.
Though a problem this time didn't affect calibration measurement fortunately, there was some possibility to be required a re-measurement in the worst case.
If sufficient checks were done, we was probably able to find a problem in ~2hrs. earlier.

4) Request state of IFO guardian was changed in mistake (though it's not directly related to the issue in this time).
=> A request to the interferometer must be done via LSC_LOCK guardian.
There is no work to require the manual operation of IFO guardian except for a bug fix of IFO guardian itself.