If this is really what you are experiencing, it might be simpler and quicker to fix the wire than wait for a new library.
There's no point in blaming the library, it has worked too well for everybody else. There's not much in blaming the sensors either. They are very reliable, and (trust me on this) they can take a lot of abuse. I have had three of them giving readings at one second intervals more or less continuously for over two years, without the faintest suggestion of a problem.
This tends to point the bone at your code. I said before that I assumed yours was derived from the same tutorial as mine. It isn't, but you might look there.
I don't really see how that can be a problem with the wire itself : if it was damaged or not properly plugged, it would not work at all, would it ? Instead of that, both sensors works perfectly fine nearly all the time, but return wrong values sometimes... However, here is a new wire-related lead : is it possible that the power drainage due to the activation of the heater, or the proximity between the data wire and 220V cables can disturb the data sent by the sensors ?
Regarding the code, I seem to have the same as everyone else : apart from the difference that I use 2 sensors instead of one, I have exactly the same code as Wildbill (thanks, Wildbill !)...
Never wait for a patch if you need it ... just apply it yourself (after making a copy of the original code of course)
Even if I can understand most of the advice and code given here, I don't know if I can mess with some library code ! But maybe it's quite simple ?
Concerning all the verbose Serial print function calls, it is my understanding that when the linker sets up the code and ram images for your sketch, the default behaviour is that effectively variables are created in the ram space which hold all of the text strings that you are printing there. It does this unless you specifically force it not to, by putting those text strings into the flash memory where the machine code of your program goes. On the arduino, most users run out of ram before they run out of flash space for their program. That is why there are so many threads explaining how to use progmem and F ( ) macros to put the text into the flash memory.
I've logged all sensors data since 1pm yesterday, and the available RAM stayed at exactly 822... Sadly, the computer manage to lose the serial connection during the night ! So, even if the system seems to have worked fine for all the night, I don't have any trace of it after around 1am :/ Maybe the serial port is not made to do any "long" data logging after all...