Calibration that do change without correspondance to run starts and ends

Suppose we have calibrations that really do change with out correspondance to runs. We no longer can apply our restriction on the database! Can we still handle them?

For example as in Figure 10.

  

Figure 11:

The above case has calibration that change rarely. Thanks to our restriction on queries, we have no problem with analysing run₂ and run₄, but run₁ and run₃ have a problem.

Another example is in Figure 11.

  

Figure:

Here the calibrations change faster than run numbers and we have a problem similar to the one in the previous example with run₁ and run₃.

As we have already talked about it, this could be handled by making database queries on an event-by-event basis. If such getData() calls have $t_{start} \equiv t_{end}$, the queries cannot be bad by definition bacause they cannot overlap two validity intervals. Since one does not want to make database accesses for each event, the database manager could pre-fetch all the calibrations $cal_1 \ldots cal_4$ nto memory and when getData() is called for each event, search for the right calibrations in this memory cache and return the right data without any database accesses.