[Backdoors are hard to find.] At this point with the exiting statement of the developers only a fool would trust Truecrypt with anything important.
Let's see: only a fool trusts things that actively lose data. (ie, bitrot, or email systems used by important people. If it's important, have 2+ independent copies)
So let's posit that TC is "sane", that it doesn't actively corrupt your data (Actual disk bitrot is another matter.)
Is it secure? (Ignoring keyloggers, CPU tampering, OS-file I/O interception, not to mention on-bus DMA controllers that have direct access to physical memory, and other out of band things? You could argue they need to detect this but they aren't an A/V vendor and you do halfway have to trust your hardware. Oh, visit CC PIN hacking via a IR camera to see your hardware "betray" you.)
Well, given a correct encryption key, things work correctly; given seemingly any incorrect key, things don't -- a very good start. So they need to protect the working in-memory key (because it's game-over if not.) They erase it if enough idle time has passed and try to keep it from being swapped out to disk. Process memory isolation is great, but in both cases the OS itself can do whatever it wants. So you have to trust the OS, at least a bit.
So, what everybody actually means: is the encryption secure? Can someone who doesn't know my password read my data due to stupid password handling, bad encryption routine choices (ROT-26), or leaky code of good routines? (Say perfect AES file encryption, but the unencrypted source file moved to the recycle bin, never mind about any corruptible buffer or stack overflows. [That's an example; TC doesn't encrypt single files.] ) Are there password collisions, ie password are actually case-insenstive? or silently truncated after 2 characters?
I suspect that you're (humans) the weakest link because of the XKCD wrench, an easily guessed password, or your likes/habits that could lead to your password. If you can't type your password it's not going to work, and you have to remember how to type it.
It seems to boil down to do you trust the vendor to act in good faith every step of the way? Let's see: -anonymous vendor, +access to source code that compiles to the released binary, +routine usage that makes sense, +updates over time, -weird final message. Personally, i trust them more than MS's native BitLocker, which is sane but has a (understandable) business-released AD key recovery function. (It's not your data but the companies, and they have keys to continue read it.) But is BL actually secure? Dunno, can't tell; we have to trust MS completely on that.
If it (TC v7.1) was good to use the day before sunset, it was good to the use day after too, until known problems arise or non-OS support kills it. But YMMV -- trust whom you see fit. So being curious: what are you using, if not TC?