Thanks, you're right - it seems OK in Pronterface, so I can use that for hand-tuning the heater. Strange that only M307 seems to be affected, perhaps something to do with Octoprint's temperature polling.

If I enter M307 (using the Octoprint terminal) to either interrogate or change the heater model, RRF responds with the appropriate output but then becomes unresponsive for approximately 2 minutes 10 seconds before returning to normal. I'm fairly sure this didn't happen in any of the previous releases of RRF. Does this happen in the Duet build as well, or just RADDS?

Here's the output of Octoprint's terminal (including routine temperature polling):

Recv: ok T:159.6 /0.0 B:24.7 /0.0
Send: M105
Recv: ok T:155.6 /0.0 B:24.7 /0.0
Send: M307 H1
Recv: Heater 1 model: gain 374.7, time constant 134.0, dead time 6.2, max PWM 1.00, calibration voltage 0.0, mode PID, inverted no, frequency default
Recv: Computed PID parameters for setpoint change: P10.4, I0.360, D44.7
Send: M105
Recv: Computed PID parameters for load change: P10.ok T:151.7 /0.0 B:24.7 /0.0
[RRF hangs for 2 min 10 sec at this point]
Send: M105
Recv: ok T:85.7 /0.0 B:24.4 /0.0

Having recently come across this thread I tried changing from M558 P4 connected via PWM3 to M558 P5 connected via AUX ADC. I'm glad I did! For me it made a huge difference. With P4 I struggled to achieve multipoint probing with a 0.04mm threshold and up to 6 repeats, but now with P5 I can do it with 0.01mm with usually no repeats. G32 calibration now reports a deviation of around 0.01mm, whereas before I might get 0.045mm at best. I'm using a very sensitive mechanical nozzle probe built into an effector I'm developing, consisting of a bolt-and-washer circuit breaker with adjustable sensitivity (will upload to Thingiverse in due course). I can't say whether changing from P4 to P5 would benefit the Smart Effector but it's definitely worth a try.