Cached completion proposal applies wrong text edit

[BoykoAlex]

Suppose user invokes content assist at position x, looks at proposals dismissed them (e.g. press Esc ) and then invokes content assist at the same position x. Completion proposals in this case of repeated content assist invokation at the same position are not coming from LS but come from some Atom cache (as there are no messages between client and server at this moment)
There is a case with a completion proposal that applies multiline text edit when that text edit results in a different text, wrong text.

Install concouse-pipeline-yaml atom extension (version 0.4.0 is the latest at this moment)
Create file pipeline.yml and open it. The grammar in the bottom-right corner should read Concourse-Pipeline-YAML
Enter content:

resources:
- name: Resource1
  type: 

Then invoke content assist where `<^> is located:

resources:
- name: Resource1
  type: <^>

Select git proposal for example. The result:

resources:
- name: Resource1
  type: git
  source:
    uri: 

The result is expected and is good.

Now undo the edit (Cmd-Z or Ctrl-Z) and invoke content assist at the same location as above. Select the same git proposal. It is important to pick one of the previously selected proposals. The result is:

resources:
- name: Resource1
  typegit
  source:
    uri: 

Where did the : go?

Note that if content assist is invoked at any other place and then invoked again at the spot above it works fine. Seems like a different edit/proposals are cached.

Seems to be related to #218 where there is no proposal shown for value with no prefix. Here the proposals are show, but they cached with an issue similar to #218.
This issue is a complete mystery

[UziTech]

Development of atom-languageclient has officially moved to https://github.com/atom-ide-community/atom-languageclient 🎉

If this is still an issue please consider opening an issue on that repo.