Test results:
//-added LPFs to SPI MOSI/CLK (220 Ohm + 10 nF) -> Tau = 2.2 us | Result: rise time too slow
//-added LPFs to SPI MOSI/CLK (041 Ohm + 10 nF) -> Tau = 0.4 us | Result: rise time borderline
//-added LPFs to SPI MOSI/CLK (020 Ohm + 10 nF) -> Tau = 0.2 us | Result: rise time acceptable
//MISO is harder due to LTC6820's open drain output. 
//-R81 to 1k7, LPF is (20 Ohms +  10 nF) -> Tau = 200 ns | Result: rising edge (pullup) doesn't source current fast enough
//-R81 to 1k0, LPF is (20 Ohms + 250 pF) -> Tau = 005 ns | Result: looks good on scope.
//A more robust method is to buffer MISO near LTC6820 (e.g. opamp).
-All LPFs (MISO/MOSI/CLK changed to (1 kOhm + 20 pF) -> Tau = 20 ns

Overall SPI deadbugging result after test drive: Absolutely no improvement to the error rate.  If anything it's slightly worse (possibly because I slowed the clock down, hence more opportunity for periodic aggressor to corrupt data).

Future SPI changes:
-Increase SPI clock rate to minimize crossover with periodic aggressor signal:
//SPI_CLOCK_DIV16: SPI CLK/MOSI look great.  MISO rising edge too slow... however, no CRC errors sitting on bench.
//SPI_CLOCK_DIV32: SPI CLK/MOSI/MISO look great.
SPI_CLOCK_DIV64: SPI CLK/MOSI/MISO look great.

Conclusion:
Change SPI to 20ish pF with 20 Ohm resistor... that'll let us add LPF if needed

/////////////////////////////////////////////////////////////////////////////////

Solution:
LPF Fc=16 MHz:
(100 Ohm + 100 pF)

Removed SPI from BCM Connector.  Added 0.1" SIP instead