I brought the furnace up again in winter of 2016, made a bunch of material for casting (cullet, cane, shard, billet), and finished some additional testing on the batch and color compatibility. I'll post some notes about compatibility separately. In short, its possible to melt batch glass with very little variation in compatibility from batch to batch, provided that the formula doesn't deviate. Compatibility is assessed qualitatively, with polarized light. Typically, I'll make a cookie (a disk) from two samples of cullet by melting them together in the glory hole, being careful not to swirl them. After annealing, I look at their interface under polarized light for evidence of stress. This method is quick, efficient, and reproducible. I also further reduced the fining agent to 0.02%, and found a practical threshold of around 0.01%. However, a formula without any fining agents may yet be realized. The seeds are very small, and tend to disappear over the course a day, which is completely acceptable for my needs. The formula was adjusted slightly to eliminate one of its more minor components, with no noticeable effect on compatibility or other practical characteristics (melt profile, working time, annealing profile, etc). I've long had a hunch that a melt "ages" over time. I did not find this to be the case however, at least over the course of a week. Samples from the first day were highly compatible with samples from the last, and indistinguishable. The cords at the very bottom of the pot likely arise from its slow and steady dissolution. I've started scooping the last 8 - 10 kg of glass out before each new batch cycle, as it does affect slightly affect the new melt. I love this glass, the formula, and the lack of variation. I love the fact that I can make it inexpensively and consistently, and that I'm not reliant on glass companies to manufacture it for me. I can hardly wait to kilncast a bunch of stuff with it over coming months.