I just bought an Osram T5 batten 28w and I thought it was instant start so I used 250v from the variac in hopes to heat the electrodes faster during starting.
On instant start ballasts the electrodes are heated up by the capacitor resonance current, which depend mainly on the voltage across the capacitor, so across the lamp. As the circuit operate in (rather high Q) resonator mode, the voltage depend mainly on the lamp glow discharge voltage, not on the mains input voltage. Consequently the same is valid for the heating current - does not depend on the mains voltage.
So increasing the ballast input voltage does not help, it may only cause troubles, mainly with the programmed start ballasts (even the simplest versions using the PTC):
The preheating current is way too high and as the filaments have shorter time constant than the PTC, it mean they would be overheating.
With the open loop frequency controlled designs it will cause higher voltage across the tube during the preheat, what could lead to premature ignition.
The most intelligent ballasts contain an overvoltage protection, what would cause the ballast to shut down, when the input voltage goes above the rated tolerance.
And if the ballast include an active PFC, it will not cause any difference, as the PFC (for "230V" mains ballasts) regulate it's DC output to about 400V regardless of the actual input. With marginally designed regulation loop frequency compensation it could cause the PFC to be unstable, causing high over- and mainly undershoots, with consequent erratic ballast behavior.
@dor: Even the simple PTC controlled ballast is qualified as a "programmed start", as with normal use cycles the PTC have enough time to cool down. The problems may arise only on motion sensors or similar simple controllers, what tend to cycle the lamp very frequently...