There's some directly relevant literature that's worth a look:
Cao Y., Liu W., Sauer P. E., Wang Z. and Li Z. (2012) An evaluation of alumina reaction tube conditioning for high-precision 2H/1H isotope measurements via gas chromatography/thermal conversion/isotope ratio mass spectrometry. Rapid Commun. Mass Spectrom. 26, 2577–83. doi: 10.1002/rcm.6378
I have monkeyed with several different procedures as well and have found that daily injections of hexanes (as Marilyn describes) provide adequate conditioning for the compounds I analyze. I inject sub-microliter quantities (0.1 - 0.5) and do not notice long-term memory effect.
This is counter to what Cao suggests (methane), but it's informed by practicalities on our end. Our unit is not configured for a methane line, for starters, though this wouldn't be hard to change. Injection of hexanes allows for some first order error checking of the GC-side of thing -- retention time and voltage response of the gas. I have also experimented with reactor temperature and found that the best isotope data comes from lower reactor temperatures (1400 C), and we try not to move it once it reaches temperature. Methane is harder to thermally convert than the n-alkanes, so conditioning at 1400 is a less efficient process. That's a minor concern, because One can simply let the methane flow longer. But, since the hexanes have some benefits and requires essentially no extra effort, my view has been, why bother?
Department of Earth and Atmospheric Sciences