After I started to use a USB pendrive to boot TC I've also noted the "... Assuming drive cache: write through" kernel message showing up during the boot process and triggered by this thread I did a bit of digging around:
(1) This message can also be seen when booting TC from a CD-ROM with any USB storage attached. It did not matter whether I used an USB pendrive or an USB hard disk and also the presence or absence of a 'tce' directory did not make any difference: the message was always showing up (and can more easily be seen when using the 'pause' boot code). I even went as far as using a FC14 live boot CD-ROM and found the same (when replacing the boot codes 'quiet rhgb rd_NO_LUKS rd_NO_MD rd_NO_DM' with 'quiet 3').
(2) The message seems to be generated by the following line of the kernel (in function 'sd_read_cache_type()' found in 'drivers/scsi/sd.c'):
sd_printk(KERN_ERR, sdkp, "Assuming drive cache: write through\n");
And it seems that due to the fact that 'KERN_ERR' is loglevel 3 the printing to the console can be suppressed with the 'loglevel=N' boot code (with N <= 3). Nevertheless the messages will be included in the 'dmesg' output.
(3) Whilst I personally think that using a default branch in said function should not lead to a 'KERN_ERR' message (but rather a 'KERN_NOTICE' one) I'd like to assume that the kernel developers have made this choice for a reason. Nevertheless due to the open source nature of the kernel one could implement such a change and compile ones own kernel.
(4) The other message is probably more "real": I did a bit of a google search with the (assumed) device ID (i.e. 058F:6364) and could not quite work out whether this
bug report is related to your issue. In any case it seems to involve a certain flash card reader, so you might want to do some tests (e.g. disconnect the device) to find out to which degree you are really affected. Also a closer study of the "dmesg neighbourhood" of that message might provide some hint.