Just a test, 5 times slomo, stabilized rendered with optical flow to 50p (with DR). Means 144 frames became 720 frames.

So sad that I cannot render this project once if I want to have 25p and I must rerender.

Why 25p, because it looks more cinematic?

One more tip I can offer: Use Shutterencoder to conform the 50p Vegas output to 25p by dropping:

As I mentioned before, I use 25p in final project, because I want a few raw videoclips/events slowed down 0.5x and other a few videoclips/events to keep normal speed (not slowed down), so my project is a mix of slowed and not slowed down videoclips.
Thank you @3POINT for showing optical flow possibilities. It looks great.

@marcinzm I showed that for 0.5x slowmotion, there's no need to go to 25p. You can stay at 50p. My example of 0.2x slowmotion is also rendered at 50p.

@marcinzm The choice of filming in 50P and delivering in 25P is entirely justified if you want pristine slow downs.
On the other hand there are drawbacks, speech and sound is one issue, and there is less light available when filming in 50P. Quality may also suffer if the bitrate is capped in camera (there is probably more bandwidth available filming in 25P in 50P capable cameras). You also loose the fast motion scene advantage.
I film in 25p 4K to deliver in 25P FHD mostly because I film a lot handheld 600m which need an additional crop for birding and wildlife filming so I can have about 1200mm zoom in full HD delivery. And using Mercalli helps a lot with that and or has head room for extra cropping due to stabilization which uses cropping anyway.

I film in 25p 4K to deliver in 25P FHD mostly because I film a lot handheld 600m which need an additional crop for birding and wildlife filming so I can have about 1200mm zoom in full HD delivery.

You still can deliver in 4k, also when you add some digital zoom to fake 1200mm, no need to downscale to FHD. The scenes you add digital zoom will look the same in 4k as they will look in FHD. But the 4k scenes without digital zoom will look worse in FHD than in 4k.

I did some tests and I must say that I am forced to return to Prodad Mercalli SAL (Stand Alone Version). I noticed that if I shorten videclip in Mercalli OFX 6 to 1 seconds I have still shaky footage in final render file even if I set 50p in Project settings and render video to 50p :((( If I would leave i.e. 3 seconds (because original file has 3 seconds) I would able to use OFX 6, but I think that Prodad Mercalli SAL could do it faster than while rendering to 50p in Vegas Pro. I used to use SAL version and Prores video files settings.
Do you have the same opinion as me?

@marcinzm  ... I still use Merc 5 SAL a lot (along with a couple of other stabilisers such as Resolve and Mocha Pro depending on the motion involved) but rarely use Merc 6 SAL because 5 SAL provides much more user control with its camera presets (e.g. Universal, Glide, Rock-Steady).

BTW, have you tried using Vegas Pro 23's native Video Stabilization plugin? I just tested it on your video and it successfully rendered from 50 fps to 25 fps on VP's timeline - and it seems (IMO) to give a slightly smoother stabilisation than does Merc OFX 6. With VP's stabiliser, I used just the bog standard 'Basic' setting without any adjustments.

I did some tests and I must say that I am forced to return to Prodad Mercalli SAL (Stand Alone Version). I noticed that if I shorten videclip in Mercalli OFX 6 to 1 seconds I have still shaky footage in final render file even if I set 50p in Project settings and render video to 50p

Strange does M6 OFX works different than M6 OFX trial? I shortened your clip to 1 sec and stabilised it with M6 OFX trial as shown a few posts earlier.

I vaguely remember that when I used extremely short clips once (like around 1 second or less) they were not properly handled by Mercalli.

@3POINT But it is weird, that preview window shows stabilized video properly, but only after render it looks shaky, even if I set 50p in project properties and render to 50p :(((

I film in 25p 4K to deliver in 25P FHD mostly because I film a lot handheld 600m which need an additional crop for birding and wildlife filming so I can have about 1200mm zoom in full HD delivery.

You still can deliver in 4k, also when you add some digital zoom to fake 1200mm, no need to downscale to FHD. The scenes you add digital zoom will look the same in 4k as they will look in FHD. But the 4k scenes without digital zoom will look worse in FHD than in 4k.

@3POINT  First of all, I deliver FHD for 3 reasons,

1. First, I just simply still have a big fabulous last generation FHD Panasonic plasma TV. No reason to upgrade since broadcast TV in Belgium is still not 4K.
2. Second reason, as I mentioned before, having a 4K source allows me to crop to FHD on the subject, not the whole picture. This has the effect of zooming in (albeit without change the perspective or "bokeh"). And if you deliver in FHD you keep all recorded pixels 1920x1080 in that crop so no quality is lost if cropping and delivering in FHD from a 4K source. It is as if filming in FHD with an optical lens of 1200m (as I have an optical lens of 600mm). So basically it mimics a 2x zoom effect without loss of quality. Cropping 4K footage to a Full HD () as it utilizes the high-resolution source to fill the frame. This provides a, high-quality digital punch-in for re-framing or creating closer shots in post-production, often called "lossless" digital zoom. 
3. Back on the poster subject, Mercalli, or any other stabilization software, reframing and cropping gives me headroom to crop in to cut out black borders generated by stabilization software or correct tilted horizons. I assume that Mercalli only uses the cropped part if placed after pan/crop (I hope that is the case, as I am only interested in fixing camera movement on the subject...

I vaguely remember that when I used extremely short clips once (like around 1 second or less) they were not properly handled by Mercalli.

I really ask myself does it make sense to show a stabilized clip shorter than 1 second?

@3POINT But it is weird, that preview window shows stabilized video properly, but only after render it looks shaky, even if I set 50p in project properties and render to 50p :(((

I cannot replicate, a one second rendered part of your clip looks to me well stabilized with M6OFX trial:

It looks not 100% perfect here because videos uploaded to the Vegasforum a re-encoded to 30fps. But the original 50 fps output of Vegas looks good.

Just to make sure, this is my approach:

@3POINT But it is weird, that preview window shows stabilized video properly, but only after render it looks shaky, even if I set 50p in project properties and render to 50p :(((

@marcinzm

The only thing I can think off is that the rendering is taking a corrupt (or a wrong) pre-rendered file (like proxies) to generated the output file.
Here are some theories/things to try out:

1. I do not know if you are using proxies (automatically generated by a setting in Vegas - or manually requested), perhaps Mercalli is only active on the proxies, or on the source file... You could try to disable proxies and see if that solves your issue.
2. disable dynamic ram preview max (put it to 0)
3. file I/o preference: enable legacy AVC decoding
4. decrease the number of rendering threads to 1 only
5. before rendering, quit Vegas, then start Vegas again, do not do edit nor play anything on the timeline, just start render.

The 25p workflow and a comparison between M6 OFX trial and Vegas stabilization:

My conclusion; I'm glad I did not spend the money for buying M6 OFX.

The 25p workflow and a comparison between M6 OFX trial and Vegas stabilization:

That's much the same as I found - the Vegas Stabilization FX provided the better result.

When I did my earlier testing, I was using the whole just less than 3 secs video. I have since tried Merc 6 OFX on a 1 sec sample in VP23 and agree that stab works in preview but not in the render (50 to 50).

As a matter of interest, I used stab in both Merc 5 & 6 SALs (default settings in both) on the first 1 sec selection of the OP's video and the stab render worked in both cases. The best stab result IMO was the render from Merc 6 SAL - it was noticeably better.

The same 25p workflow with M6 OFX trial and Davinci Resolve stabilization:

Also here M6 OFX produces worser results, my conclusion it's not a Vegas only issue, M6 OFX is not worth the money.

Yes, I confirm your conslusion.
The appropriate results (on other files, but overall) I only got after using M6 SAL.

I also tested Topaz Video stabilization and it gave me good results but stabilization processing run slower than in M6 SAL.

So in my opinion the best option is M6 SAL.
Can you also share the same desktop screen videos as you recordered, but while using M5 OFX?

Maybe M5 OFX will work better than M6 OFX?

I am shocked :) I have just used M5 OFX and it worked better than M6 OFX :)

I didn't even set 50p in project settings, I used 25p and output rendered file was stabilized well.

So my conclusion is not to use M6 OFX but M5 OFX :)

I have just post a ticket to Prodad support to solve this issue in OFX 6.

I don't know how OFX 6 is better than OFX 5?

Can you also share the same desktop screen videos as you recordered, but while using M5 OFX?

I do not have M5 OFX (does that exist?), I only have M4 Plugin and M6 OFX trial.

There is a Mercalli V5 Suite for Vegas, but no OFX.