Good to hear this is working now!
Using SIO.SYS /A in config.sys (To access OS based Hias SIO patch) now lets me access normal ATR based RespeQt ATR media at divisor 0 reliably. However, PCLINK.SYS access only operates at 19200 in this mode.
My OS SIO patch only accelerates D1:-D8:, so PCLINK will run at standard speed - so the result is to be expected. If you have a U1MB you could try using it's PBI SIO driver, IIRC this will accelerate PCLINK as well.
I had a look at the SIO connection part of the 1050 schematics:
1050-sio.png 217.49KB 1 downloads
The caps C56,C57,C58 (which are recommended to be clipped off) create a path from SIO data in to ground (C61 which is also recommeded to be removed is behind the CA3086's transistors so won't affect the SIO bus directly). With a 4k7 resistor in series and a total capacitance of about 19pF that won't be a huge load - though it'll still affect the SIO bus even if the drive is powered off.
I also measured the parasitic capacitance of a SIO cable (between SIO data in/out lines and GND lines) and got about 110pF.
So the worst "offender" seems to be the 1nF caps in the Atari, followed by SIO cable, then parasitic capacitance in the SIO device (like the caps in the 1050), each about one order of magnitude apart.
The additional pull-up may help a bit in extreme cases (caps in the Atari plus a few SIO devices connected), but I'd guess a similar pull-up would be needed on SIO data out as well.