That's not unexpected. The constant 1.5V (maybe slightly higher) output voltage is too high for the bulb. It's probably a 1.2V rated bulb. In a normal battery the 1.5V drops pretty quickly hence not a problem.
Another way the Batteriser can be detriment instead of a benefit.
Sure, lightbulbs suffer greatly from overvoltage, nothing new here, but the ones I used were rated at 1V5 and the batterizer is outputting 1.51V RMS. So that SHOULD be ok, at least if those bulbs are definitely 1V5 rated and not "to use with 1V5 batteries", as I can only read the marking on the bulb itself which states 1,5V.
Even if they were not designed to run at a constant 1V5, they shouldn't just fail after around half an hour, should they? Also, I had two batches, three came with the package from frank, and just for comparison I bought another set of four at the local conrad store.
Either way, just for curiosity I digged out the scope (as I said, lab is being restructured right now, so just a few quick measurements...) and looked at what we have here.
First, no load connected. The Batterizer is switching between battery voltage (~1.48V) and around ~1.64V with around 10 kHz.
another one with bandwidth limitation
Switched on, measured directly at the load, connected with short cables: around 200mV of "noise"
Let's zoom in a bit: switching between battery voltage, which breaks down to 1.39V under load and 1.61V with a frequency of around 40 kHz.
there are tiny spikes reaching up to 1.67V
A bit closer to measure the frequency: 40.48 kHz.
A last one without the cursors, just for curiosity. Interesting waveform with nice tiny spikes at the rising edge.