Update: I ran the 9RGB version for 10 hours yesterday. This uses the same 9 bit base sourcing multiplexed 3-sinks transistors. PIC was cool as a cucumber with no noticable temporature change or even slight warming! That tells me 15RGB version based on the original posted design should work without any problems. Although suggestion of spreading the source pins across the 2 ports is a brilliant safeguard.
Based on the PIC datasheet electrical specifications: I did notice 200mA maximum combined port currents (PortA + PortB). I did not see any indication that each port current was limited to 100mA. Something tells me that the current limitation is not necessarily related to the ports, but rather limitation of the chip in its entirety accomodating certain supply current thru it (VDD, VSS). This could be due to internal wiring (semiconductor masking).
Further testing: Although my color motion version does not use white or pastels (All 3 bits on R,G,B), I intensionally forced white color across all RGB LED's. For short period of time chip got warm and for a longer period of time the chip got even warmer. This confirms the current limitation of the PIC and contineous over-driving the chip beyond its capacity may result in failure.
Based on calculation the currents during normal operation of the color motion, with two color bits used (RG, or GB, or BR): 6 pins used on the source side, total current draw is 150mA (6x25) which is within the tolerance of the specification confirming the above findings.