An exact centre tap is handy if that's what you need. But I've come across designs where the 50Hz hum from the heaters (adjustable via the humdinger) is used to null out 50Hz leaking in from other sources e.g. magnetic flux from the mains transformer, or even ground loops. For that you need to be able to offset the heater contribution away from the centre position, where all it would really be doing would be minimising itself.
The problem then arises with higher harmonics - commonly, for symmetry reasons, odd ones. Now we can set the humdinger to null the 50Hz component, but there may well be 150Hz and 250Hz ones too. The heater waveform itself may (roughly) replicate the mains waveform, which these days can be very far from a pure 50Hz sinusoid. Flux leakage from a transformer can have its waveform strongly modified by the iron's nonlinear B-H curve. Contact resistances which convert ground loop currents into noise voltages can be nonlinear too. So we end up with a number of hum sources each of which has a different waveform and harmonic spectrum. The question becomes whether to null the overall 50Hz component, which might have the largest amplitude, or, say, the 150Hz one, to which the ear might be 16-17dB more sensitive
https://en.wikipedia.org/wiki/A-weighting. This is when the alternative approach of eliminating the hum at its sources by using DC heaters, well-behaved and positioned mains transformers, ground loop current reduction etc really pays dividends.
Cheers,
GJ