I have a BF561 based system (see http://unaligned.org/bigcam/dw.php for schematics) that contains a -6A version part. The chip has been running at 600MHz CCLK, 120MHz SCLK for a number of days, with a fixed external regulator supplying 1.33V VDDINT and 3.3V VDDEXT, in open air without a heatsink. There were no problems with the part at that point.
According to EE-293, the estimated power for the part would be in the 1.5W range. Since the Th_ja for the BC-256-4 package is quoted at 18.1 C/W in still air (0 m/s), no heatsink was expected to be required as the chip was running in open, room-temperature air on my desk as the junction temperature would be in the 50-60 C range. The case temperature would agree with that since it was finger-warm but not burning (I know that's not very scientific, but that usually seems to mean 45-55 C for me).
However, after this time, attempting to run the part at 600MHz CCLK will result in an unstable system (spontaneous reboots). In fact, I had to back off the clock to 400MHz to make it stable. The core voltage is still 1.33V and quite stable, and I can still run the SCLK / SDRAM at 133MHz with no issue whatsoever, and running it at 100MHz does not help with stability issues either. Only CCLK needs to be de-rated.
Has anyone observed that kind of degradation in BF561 at prolonged runtimes at full rated speed?