The converter's "collateral price" in this case is already a 3.5 day median

My proposal above was the use the most recent price. Using the 3.5 day median isn't great because it may be a lot higher than the current price at the time the conversion is initiated just due to market fluctuations, which may result in very low collateral, even without any manipulation. (This was the motivation for your use of minimum with a recent price; the same logic applies here.)

And a minimum over a 3.5 day period, or even a 24 hour period could potentially be very low, especially if there was one malicious witness feeding a low price. A single witness could effectively negate use of the conversion mechanism in this case

I meant minimum of the hourly medians, so any individual witness with a bad price would be filtered out. This is already the current mechanism in the code.

If the price has really risen a lot over 1-3.5 days (say 2-3x), then I would say there is enough volatility to be "suspicious" (not suspicious in the sense of manipulation, necessarily, just that it may not last), and not give credit for that new, higher price right away, in terms of collateral required.

Overall I'm not very concerned with requiring more collateral. Locking up HIVE for a short time (1-3.5 days) is a very small cost if there is direct profit to be made from the conversion itself (particularly if the profit is large). Arguably, giving people a reason to hold and lock up more HIVE could even be good, but that's not my primary motivation for suggesting a more conservative collateral calculation in this case.