IMHO, no one isolator (except on a transformer, and even then) has low enough jitter to pass a master clock from the interface to the DAC.
The oscillator should be located close to the DAC chip, and from there it should go to the interface through an isolator - in this case, the jitter of the isolator does not affect anything. For DACs sensitive not for MCLK only, other signals may be re-clocked.
Well that's how PA works, not a failure. PA reserves the devices it is allowed to access. On a desktop linux all applications go through PA, by default. If you need an application to go directly through alsa for some device, you have to tell PA to ignore that specific device, e.g. https://askubuntu.com/questions/666...rtain-device-from-being-grabbed-by-pulseaudio
I agree. On this particular board MCK should bypass the isolator. But still ISO7240 with up to 1ns jitter is not something I would use for I2S signals either.
ISO are not so bad, they are less noisy than ADUM.
I use SI86xx for I2S, very quiet, and I don't care about their jitter.
I never studied the details of device handling in PA. It may try to access the devices periodically to check if they still exist and are in a working state. For the default device (often the last one connected, unless specified otherwise) it may open it and keep it open (while playing zeros) for longer time, to avoid possible clicks in amp on/off, to allow smooth mixing in new streams, etc.
The safe rule is to disable a device in PA if you want to use it outside of PA. The setting is done only once, the config is persistent, the device can be disconnected/reconnected anytime afterwards.