Calo Delta counters are unstable

Currently we have a special treatment in multithreaded test validation w.r.t. some counters (see here)

sensitivities = { 'FutureEcalShare...' : {'Delta(Z) Outer':0.001, 'Delta(Z) Inner':0.001, 'Delta(Z)':0.001 } }
validateWithReference(preproc = preprocessor, sensitivities=sensitivities)

This effectively hides the problem from the nightlies by allowing for big variations. Nevertheless, the underlying problem should be fixed (it is possible that the condition number for these sums is very high and they are thus not a good thing to look at anyway).

In addition, another calo alg is showing an unstable counter (although this one not in multithreaded mode):

Different content of Counters for algo CaloTrackBasedElectronShowerAlg_...1
    (CaloTrackBasedElectronShowerAlg_...1 ref) average DLL | 4729 | -325.8811 | -0.068911 | 0.096923 | -1.4142 | 0.23314
    (CaloTrackBasedElectronShowerAlg_...1 new) average DLL | 4729 | -325.967 | -0.068929 | 0.096903 | -1.4142 | 0.23314

See Moore!1279 (comment 5371714) as well