Gubi,
Yes, I understand that indi_synscan_telescope is just passing orders from the client, with a bit of minor translation/formatting to the SynScan app (Android, in my case) and that all the "work" and heavy calculations are done in the app and passed to the mount. This compared with the indi_skywatcherAltAzMount, that does both the client communications and calculations as well as communications to the mount. A much easier job for the indi_synscan_telescope driver.
Let me mention, while I am thinking of it, there is a small bug in indi_synscan_telescope regarding the Site properties. It can't impact much, but as a matter of "neatness," location elevation will not set and reverts to 0. One may even alter the .xml and place an elevation number there, but it will still revert to 0. Not a big issue, but worth fixing.
Plate solving failure, in general, has only a little to do with the driver. I have grab some of the temp images from failed solve attempts and usually find the failure due to bad image quality, causing a time-out. This, usually, is due to poor tracking. If the scope isn't still when it tries to image, you'll get a fail. The problem is that there is sometimes a bit of a delay after a GOTO while the scope settles. There probably needs to be a configurable delay after GOTO and between solves to prevent a moving scope from being the reason for the failure. The other major causes of solve failure seem to be time-outs. These probably aren't driver related and have more to do with the image specification and the parameters for astrometry.net (local). This incluldes selection of indexes. Too many or wrong indexes will cause timeouts that shouldn't happen. Also, as you noted, a blind solve must not have a radius limit. Frankly, the radius limit for close range goto solves should probably be very visible as a configurable option.
I've been recently getting as up-t--speed as I can on general plate-solving. My current mode of operation is to startup the scope level and North. Use the Synscan app (without alignment) and slew to "some target." It doesn't matter which, but I typically pick some bright star. Then I manually take an image and manually blind plate solve with local astrometry.net conver the RA/DEC solved results to Alt/Az and compare it to where the scope thinks it is. Manually move the scope (with Synscan App or SkySafari->Synscan App arrow keys in the general direction of my calculated Alt/Az, then sync. And then repeat the solve and move, though I add a 30 degree radius to the astrometry.net solve-field parameters instead of blind solve. A very few hops and I'm at the target. I have to be careful to keep the hops small, or I get a sync not possible error from SkySafari. This seems to setup the alignment, as subsequent GOTO commands are more accurate.
I know that this is basically what Ekos is doing automatically. But by doing it manually, I have a greater understanding of the pieces and have taken all indi drivers out of the mix. Now, if I add a driver, I'll know the driver was responsible for any misdeeds. I did, however, get my crayons out and put together a Java app that grabs the image, calls solve-field (astrometry.net) with my parameters and does the RA/DEC to Alt/Az conversion for me, but I did do it manually first to prove the concept. So far, tracking with SkySafari running the SynScan app is the smoothest I have ever seen...much better than with the SynScan app on its own. I would love to run a capture and see what SkySafari is sending to the SynScan app, but it only works if they are both on the same tablet, so there is no network traffic to grab easily.
I am not sure about the slowness, but I note the difference in my manual process. If it doesn't do a proper sync, or the sync fails, I can see delay problems. If it doesn't reset to blind solve and keeps trying that 15 degree radius (that I believe is too small for many scopes), the it will end in a solve fail, because it is trying to solve on a radius from a place it is no longer pointing.
Try PHD2 for guiding with the indi_synscan_telescope driver. The trick is towait for calibration fo finish, then run the guiding assistant for a few minutes, stop it, then follow its recommendations. After the settings changes, it settles down nicely. Not as crisp as an polar RA/DEC mount, but certainly smooth enough for fast-imaging scopes like those you and I have.
I do have the new driver version with your additions. I'll make a screen shot, as I have a few questions. I did change the setting to Alt/Az.